|
| Greg's diary |
| January 2014 |
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Wednesday, 1 January 2014 | Dereel | Images for 1 January 2014 |
| Top of page | ||
| next day | ||
| last day |
|
Olympus “Viewer” revisited
|
Topic: photography, technology, opinion | Link here |
So was yesterday's upgrade to Olympus “Viewer” 3 worth the effort? Reading the release notes (1 line to a paragraph, of course, and requiring reformatting even under Microsoft before they can be read) suggest that there's nothing much new. In fact, most of the document is boilerplate, including a prohibition of reproduction in whole or in part—a rather silly restriction for something that's freely available on the web. But it says hardly anything about the changes, just what's on the web site. Punctuation is original, but I've fixed the markup to validate:
|
||||||
But that's yet another bug. In fact, there are some useful improvements. It's now possible to select raw “development” directly rather than via a menu uncomfortably squashed in a corner of the window. And most importantly, it's now possible to write 16 bit TIFF images with EXIF data, though they're clearly still grappling with the idea. The TIFF selection now allows you to say whether you want EXIF data or not, but they still have the old 8 bit “Exif-TIFF” selection:
|
|
||||||||
|
|
||||||||
This difference is more than cosmetic. It copies the complete EXIF data, which apparently allows enough space in the output JPEG for all of the copious maker notes. This means that the panorama problems I reported earlier are no longer an issue. On the other hand, it still forgets its raw processing settings on return to the main window.
|
Capture One revisited
|
Topic: photography, technology, opinion | Link here |
It's becoming increasingly clear that I made a mistake buying the latest version of DxO Optics “Pro”, since it no longer supports my camera hardware, and it doesn't look as if it ever will. Thus the interest in Olympus “Viewer”. But that's only part of the picture. Yes, “Viewer” can convert raw images to TIFF or JPEG and correct for lens distortion (but not for Chromatic aberration). At the moment I'm using DxO to process the output, and it's not well suited. Agreed, it's a little faster than processing the raw images, but not much.
I tried Capture One “Pro” a year ago and basically came to the conclusion that, though it had some advantages, it wasn't much use because it couldn't correct for lens distortion. Now the picture has changed, so tried it again.
First surprise: after downloading several hundred megabytes of data, it wouldn't install: “This program can only be installed on versions of Windows designed for the following processor architectures: x64 Itanium”. Maybe it was like that last time too; I was using a 64 bit evaluation copy of Microsoft “Windows” 8 at the time. Now I'm running a 32 bit version on dxo, a machine whose name is gradually becoming inappropriate. So I had to download version 6, the last that supported 32 bit Microsoft. And once again the pain of learning how to use it. It has hundreds of features, and might well be worthwhile, but somehow the mindset of the developers doesn't overlap much with mine.
|
Network speeds revisited
|
Topic: technology, opinion | Link here |
I've been dragging my feet with network speed measurements, mainly because of the pain of analysing network traces. Tried setting some tuning parameters described in this file, in the process ignoring the warning:
But kern.ipc.nmbclusters was already set to 25600, so there didn't seem to be much danger, and of course it didn't cause any problems. So I applied it to my external server as well. Throughput not improved.
Then Yvonne reported problems: her mail wasn't going out. A check showed that postfix on the external server wasn't responding. Restarted it and got the strange message:
It took a while for me to associate this with the tuning work I had done yesterday. But it proved that on this machine kern.ipc.nmbclusters was set to 8768—the default value depends on the machine memory, and this one has only 256 MB. Reset net.inet.tcp.sendspace to 65536 and all was well.
So back to thinking about the basics: I have a link with an RTT of, say, 100 ms, and a speed of 25 Mb/s. So my window size needs to be at least 2.5 Mb to saturate the link with a single TCP connection. If I'm only getting 250 kB/s, or 4 Mb/s, that suggests a window size of 400 kb. And if I run transfers in parallel, the speed should increase relatively linearly until I start running into other bottlenecks.
So I tried running with axel with concurrency between 1 and 10:
=== grog@eureka (/dev/pts/25) ~ 25 -> for i in `jot 10`; do axel -n $i -a http://octopus.com.au/speedtest100mb.bin; done
The results were interesting:
| Concurrency | Time | Speed (kB/s) | ||
| 1 | 6:23 | 265.33 | ||
| 2 | 4:57 | 342.04 | ||
| 3 | 2:43 | 620.82 | ||
| 4 | 2:51 | 591.61 | ||
| 5 | 2:11 | 775.09 | ||
| 6 | 2:03 | 821.70 | ||
| 7 | 2:15 | 748.03 | ||
| 8 | 1:33 | 1083.07 | ||
| 9 | 1:58 | 861.55 | ||
| 10 | 1:40 | 1011.93 | ||
Clearly RTT is not the limiting factor. A 25 Mb/s downlink should be capable of nearly 3 MB/s, so this is seriously suboptimal.
Went looking on the ISP sites for speed tests, and found this page on the Exetel site, which showed typical downlink speeds in the order of 5 to 6 Mb/s, worse than any I have measured elsewhere. But that's OK: at least they're going to have to admit some issues. I wonder where they are.
|
More watch setting
|
Topic: general, opinion | Link here |
Reset my watch again today. It was 5 seconds fast, a gain of 10 seconds in 27 days. Reset to -4 seconds. I should do this twice a month.
|
Kimchi soaking
|
Topic: food and drink, opinion | Link here |
My kimchi recipe calls for soaking the cabbage in 6% salt solution, 50% more in weight than the cabbage itself (typically 3 litres of water for 2 kg cabbage). On occasion that has been a problem if I don't rinse the cabbage properly after soaking. Last time I tried it with only 4% salt. That's definitely too little: the cabbage didn't soften properly. Today I've gone back to 6%, but that may not be the end of the story.
|
Spaghetti alla carbonara
|
Topic: food and drink | Link here |
We had a number of cooked noodles left over from some dinner or another, and decided to eat them today with a carbonara sauce. How? Which recipe? The one I chose was from an Italian cookbook and tasted OK, but Yvonne wasn't overly impressed. It used Parmesan cheese, which appears not to be the correct one; most recipes that I later found online suggest Pecorino. To be improved on.
| Thursday, 2 January 2014 | Dereel | Images for 2 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More network throughput investigations
|
Topic: technology, opinion | Link here |
Talking about my network throughput issues on IRC today, and I went to prove some point with axel. To my surprise, I got good throughput: 2.2 MB/s, or about 17.6 Mb/s. That was with SkyMesh, so I tried it with Exetel. 1.1 MB/s. Clearly this difference has nothing to do with the National Broadband Network. It also implies that SkyMesh also has throughput issues, just that they weren't showing very much today. Sent off a ticket to Exetel, and got a reply asking for some strange tests:
Please go through the test below and forward us the screen capture for further investigation.
http://debian.mirror.exetel.com.au/debian-cd/7.0.0/amd64/iso-cd/debian-7.0.0-amd64-CD-1.iso
http://debian.mirror.exetel.com.au/debian-cd/7.3.0/amd64/bt-cd/debian-7.3.0-amd64-lxde-CD-1.iso.torrent
http://debian.mirror.exetel.com.au/debian-cd/7.3.0/amd64/bt-cd/debian-7.3.0-amd64-netinst.iso.torrent
http://debian.mirror.exetel.com.au/debian-cd/7.3.0/amd64/bt-cd/debian-7.3.0-amd64-xfce-CD-1.iso.torrent
http://debian.mirror.exetel.com.au/debian-cd/7.3.0/amd64/bt-cd/debian-7.3.0-amd64-CD-3.iso.torrent
http://debian.mirror.exetel.com.au/debian-cd/7.3.0/amd64/bt-cd/debian-7.3.0-amd64-CD-2.iso.torrent
Download the above 6 files simultaneous and take a screen capture while the downloads are in progress.
That's interesting for a number of reasons: firstly, they're all Linux ISOs, secondly most of them are torrents, which my browsers just download as a small file. And what screen image should I show? Clearly she expects me to do this with a web browser, and not all browsers show download speeds. I'll work out how to give her what she really wants tomorrow.
|
More CaptureOne pain
|
Topic: photography, technology, opinion | Link here |
Spent some more time looking at Capture One “Pro” today. It's really hard to understand, and the documentation is nowhere near as good as that of DxO Optics “Pro”. By the end of the day I still didn't know how to process (sorry, “export”) an image. What I did find was that it creates enormous quantities of files in a subdirectory:
=== grog@eureka (/dev/pts/30) /Photos/00-Oly 3 -> l -R CaptureOne/
=== grog@eureka (/dev/pts/30) /Photos/00-Oly 4 -> du -s CaptureOne/
That last value (798) is the size of the data in megabytes! What's it for?
|
Reconsidering house layout
|
Topic: Stones Road house | Link here |
Where on the block are we going to put our new house? We had a pretty good idea, but then I reconsidered. The natural pond could be quite pretty, but it's not really in an ideal place—20 metres to the south-west would be a lot better. In any case, over today and tried to peg out the location of the house:
Back home to get a rod to dig holes in the ground, and finally ended up with pegs showing the location of the house, not easy to photograph. I put Yvonne next to each peg in turn, but by the time I got home I had forgotten which was which. There must be an easier way:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
At the end, checked our measurements. The house hasn't moved much: 4 metres north, 1.5 metres west. The big difference is the move north, which will free up the verandah from the pond.
|
Hitler was a British Agent
|
Topic: general, opinion | Link here |
Some spam I get is so funny that it's worth reading. But one I received today completely blew my mind, trying to sell a book about the ultimate conspiracy theory: “Hitler was a British Agent”. It's not even the only one: as everybody knows, Stalin was also trained in the UK. This is so silly that it's almost worth reading. But then, that's what I said about a book I inherited when I moved into our house in Rosbach, „Das dritte Reich 1918-1933“, by Gerd Rühle (Vaterländische Verlags- und Kunstanstalt GmbH, Berlin). I started reading it in the assumption that there must be something of value in there, but I was mistaken.
Jamie Fraser suggested:
i feel like there's a ton of money in you writing a book about how winston churchill was a nazi plant, in a glorious multi-decade plan to lead germany to become a power-house of a new "eurozone"
But too late. Already been done. As the book about Stalin reveals, Winston Churchill was the son of King Edward VII.
|
Which web radio channel?
|
Topic: multimedia, opinion | Link here |
Since we have fast and reliable networking, I've almost given up listening to broadcast radio, and lately we've been listening to Radio Swiss Classic, which has a remarkably varied and interesting programme. But when you think of music and small alpine states, there's another one that's much more obvious. And today was the New Year's concert of the Wiener Philharmoniker, which I've been listening to on and off for 50 years. Surely there must be some good web radio to be had?
It doesn't seem so. ÖRF doesn't have a classic channel! By chance found Radio Stephansdom, but they have advertising! Maybe it's not commercial, but it doesn't measure up to Radio Swiss Classic.
| Friday, 3 January 2014 | Dereel | Images for 3 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Goodbye Capture One, hello ACDSee
|
Topic: photography, technology, opinion | Link here |
A little more playing around with Capture One “Pro” today, but not much. Like all such software, it seems to insist on looking at everything in a directory, whether I've told it to do so or not. In my test directory I had 302 images (from last week's house photos), and it had to go and make thumbnails of every one, using another 900 GB of disk and taking 8 minutes to do so.
That's what I wrote, but it seems implausible. I suspect that I meant to write 900 MB. It's unlikely that a single “thumbnail” would consume 3 GB.
But this is release 6, and the current version has been release 7 for some time. I'm only using it because dxo, my Microsoft box, (still) has a 32 bit operating system. I don't recall having quite this much trouble with release 7, so I'll postpone further investigation until I upgrade the system.
In the meantime, went out looking for alternative software. There's so much out there. How to make a choice? Google brought up a couple of pages, including “configurator” that asks questions about what I want and then gave me a choice between only 5 programs, only 3 of which I had even heard of. Still, it was worth the trouble, and it ended up recommending ACDSee Photo Editor, which proved to be quite basic. Looking through their range, though, found ACDSee “Pro” 7, clearly superior because of the Pro and the considerably higher price, so downloaded that and tried it out.
There's something basically difficult about using any new photo processing software. Without exception they want to rearrange the way I think about file systems. It's faster than the other software, and it also understands Olympus' raw format, though it can't perform automatic distortion correction. But to its credit it makes a lot more sense than the other programs, though its help system displays text in a size that is just too small to be comfortable, and there doesn't seem to be any way to change it. Fought my way through Microsoft's Control Panel and increased the global text size, with the result that everything else was unpleasantly big, and the help didn't seem to have changed in size. Still, they have documentation online, both a 427 page manual and help texts that appear very similar to the too-small help pages.
But they still need to be read. Managed to convert a number of photos, but it's anything but optimal. If I stay with ACDSee I will have a lot of reading ahead of me.
|
OM-D E-M1 settings revisited
|
Topic: photography, opinion | Link here |
I've had my Olympus OM-D E-M1 for a month now, and I still don't have it set up the way I want. Spent a lot of time playing around with the “Myset” settings, which I'm summarizing on this page. There's so much to learn. In the middle of it, Chris Bahlo rode by on Chewey (or is that Chewie?), so I had to go out and take some photos. Still more surprises: this is the first camera I have ever seen where you see the image you have just taken in the viewfinder. For years DSLRs have displayed it on the monitor, but the viewfinder always showed the view through the lens. It's useful, maybe, but also quite irritating when you're trying to take photos in quick succession.
I've been trying for weeks to get the viewfinder to display properly in manual exposure mode even when the “exposure” is incorrect. That's typically the case when you're using flash as the main source of illumination. I've scoured the entire “manual” several times, but finally I've found it: it's a function called “Live View Boost” on the second page of menu D. You set it to “On” to get a normal display. And there's barely a mention of what it means in the manual, just this at the bottom of page 92:
If [On] is selected, priority will be given to making images clearly visible; the effects exposure compensation and other settings will not be visible in the monitor.
The function isn't new: as the name suggests, it comes from a time when “Live View” was just one option instead of the only way to view the image. I now see that the E-30 had it, and in that manual there's more of an explanation:
During live view, set [LIVE VIEW BOOST] to [ON] to make the subject easier to see without the exposure reflecting on the LCD monitor.
And then later:
During live view shooting, you can brighten the monitor for easier confirmation on the subject.
[OFF] The subject is displayed on the monitor with the brightness level that is adjusted according to the exposure being set. You can shoot while confirming through the monitor in advance to get a picture that is to your liking.
[ON] The camera automatically adjusts the brightness level and displays the subject on the monitor for easier confirmation. The effect of the exposure compensation adjustments will not be reflected on the monitor.
That's not spectacular either, of course, but the difference in detail is symptomatic of the catastrophic quality of the E-M1 instructions.
|
Jenny Jones arrives
|
Topic: general | Link here |
Jenny Jones is spending some time with Chris Bahlo for work experience that counts towards her Veterinary Medicine course. She is staying with us.
| Saturday, 4 January 2014 | Dereel | Images for 4 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More panorama refinement
|
Topic: photography, opinion | Link here |
As decided last week, decided to take this week's house photos with manual exposure and HDR. The results were mixed: some were quite good, but in some cases I just didn't have enough dynamic range. The verandah panorama was a particular problem: firstly, the 6 EV exposure difference weren't really enough for that particular view, though it worked elsewhere. And secondly, only this time, I wasn't able to join the roof images to the others, even after taking a second series of images. What's causing that?
|
Running out of resources
|
Topic: technology, photography, opinion | Link here |
Photo processing was interesting for other reasons. Everything was so slow! “Viewer” 3 took nearly a minute to process each image, speeds that I've only seen with DxO Optics “Pro”. And when I processed the output with DxO, it was also only half speed! I've seen inexplicable (to me, anyway) differences in Microsoft processing speed in the past, but with 300 photos to process, this was worth more investigation.
What really surprised me was that the Viewer batch module was using 50% of CPU time—when it should be idle. Once I stopped Viewer, DxO carried on at its normal speed, about 25 seconds per TIFF image.
Looking at the next batch of photos to put through Viewer, I discovered that the batch processing took an average of 54 seconds. Normal processing took 31 seconds. So it seems that the “batch” facility of Viewer has serious performance issues. “Don't do that, then”. But it does leave me wondering what kind of testing goes into these products.
Those weren't the only problems: the bigger photos from Olympus OM-D E-M1 lead to bigger panorama images, and the 8 GB memory of eureka is having difficulty coping. Twice I had an X crash, and on one occasion Hugin failed:
What's that? Clearly nona is too leet to tell us the real reason. But it proved to be that the intermediate TIFF files took up the remaining 20 GB of disk space on the file system. Clearly it's getting to be high time to upgrade my system to something bigger and faster.
|
Understanding E-M1 settings
|
Topic: photography, opinion | Link here |
Spent some more time refining my settings for my Olympus OM-D E-M1, in the process confusing myself mightily. I've already commented about the difficulty of accessing “Myset”, the 4 stored settings. But clearly there's something funny about the way it works. You can assign “Myset n” to a number of buttons, and if you press the button, it loads those settings. So it made sense to allocate a different Myset to the button (in my case, Fn2) for each setting: in Myset 1, assign Myset 2; in Myset 2, assign Myset 3, and so on. But somehow that doesn't seem to work, and it's really difficult to see just what settings have been loaded.
Also reading a preview of a book by Reinhard Wagner on the subject of the camera. It's not freely available, though possibly the URL is available on the German Olympus Forum. It goes into some detail about various aspects of the camera, and maybe I can learn something from it. But I have my doubts about the accuracy of some statements: it claims that the autofocus bases its assumptions on the “JPG” image. What if there isn't a JPEG? And to use the JPEG, you first need to convert, which takes time. And that's exactly what you don't have.
As a result, it says, you shouldn't use things like soft focus art filters, because it makes the focusing slower. OK, that's easy enough to test. Put on my slowest focusing lens, the Zuiko Digital 18-180mm F3.5-6.3, and tried it out. No difference that I can see, but I accidentally took a completely boring photo of a corner of my office:
|
|||||||||||||||||||||||
The first impression is that it's unsharp, but that's a feature, not a bug. In fact, I can't see any evidence of camera shake at all—and that in an image taken at a shutter speed of 0.6 seconds without any preparation. I continue to be amazed.
|
Tanya injured
|
Topic: animals | Link here |
In the late afternoon, heard a loud yelp and whingeing from Tanya. It appears that she hurt a front paw, though it's not clear how. She carried on limping more or less for the rest of the day. Hopefully it's nothing serious.
| Sunday, 5 January 2014 | Dereel | Images for 5 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Tanya still limping
|
Topic: animals | Link here |
I don't know what Tanya did to herself yesterday, but she was still limping today. If it doesn't improve by tomorrow, I suppose we should drag her off to the vet.
On the other side, she is growing by leaps and bounds. We've had her for 2 weeks now, and it seems she has doubled in size, though photos tell a different story (here 13 days ago and today):
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Still, she's growing fast.
|
No wind at all
|
Topic: general, opinion | Link here |
For weeks my weather station has reported complete calm, no wind at all. It's not surprising:
|
|||||||||||||||||||||||
Cut it free, and suddenly we have wind:
|
|||||||||||||||||||||||
The readings are low, but I don't think it was ever very accurate. I'll reconsider placement when we move into the new house.
|
cvr2 crash
|
Topic: multimedia, technology, opinion | Link here |
In mid-afternoon discovered that my recording of the news had failed: cvr2, the recording computer, had powered down. And it wouldn't come up. Dragged it into the office, where it powered up normally. And it did so again when I put it back in its cupboard. What caused that? No idea. Hopefully it was a one-off. For some reason, I've had more trouble with that machine (or the machine with that function) than with most of the others.
|
More Viewer insights
|
Topic: photography, technology, opinion | Link here |
One of the disappointments about the new Olympus “Viewer” 3 is that, although it saves the EXIF data, it appears not to save it all. In particular, the “crop factor” information that Hugin wants is not there, so once again I have to enter it manually in every panorama.
But it has a function I hadn't noticed before: “import” photos from a camera. I have never connected an Olympus camera to that machine, but today I took a memory card from Yvonne's camera and put it in there. Up pops a Viewer screen and offers to do things with the images. All I wanted was to copy them, and it could do that too, even into “folders” named after the date they were taken.
Tried it both ways: with Microsoft COPY and with Viewer. COPY copied 935 MB across the network in 2 minutes, 21 seconds. Viewer copied almost all of the files in 5 minutes, 8 seconds, more than twice the time. And it left out the video thumbnail images, probably because it didn't recognize them. On a more forgivable note, it used a different convention for the directory name: for today's date I use 20140105, while it used 2014_01_05. Nothing wrong with that in principle, but it's not my way, and I can't change it.
Once again, I'm left wondering: what use is this kind of software? Every operating system comes with basic utilities that copy files. Launching a slow behemoth to do things that aren't quite what you want seems a complete waste of time. The best thing about this functionality is that you can turn it off.
|
Another Ashampoo
|
Topic: photography, technology, opinion | Link here |
I should be trying out the various photo processing packages that I have downloaded, but somehow I couldn't face it. In the process, I asked myself what functions I really needed. It's not much: mainly perspective adjustment and cropping. Even xv can do the latter.
But wait. There's something more basic: automatic exposure correction. I've been using Ashampoo photo optimizer on most photos for years now, and it tends to improve the overall appearance of the images. What else do they have to offer? Took a look and came up with Photo Commander 11 (and not even “Pro”). It offers all the usual useless functions like backup and restore, and it's not even clear whether it can do perspective correction, but it seemed worth investigating.
Downloaded and installed: on the positive side, it includes the functionality of the optimizer and even offers panoramas. It makes more sense than many other packages, but clearly I'm going to need to read the manual. The panorama seemed to work, but I couldn't find a way to display it. I don't see myself using the panorama functionality, but if the rest is as straightforward as it seems, it might be a good choice.
| Monday, 6 January 2014 | Dereel → Cape Clear → Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Eliminating Ashampoo
|
Topic: photography, technology, opinion | Link here |
More investigation of Ashampoo Photo Commander 11 today. My fears are confirmed: it can't do perspective correction. And the real issue, clever automatic exposure adjustments, also seems to be inadequate. In fact, it doesn't do anything that GIMP can't do, costs money, and requires Microsoft to run. So, once again, it has nothing useful to offer. Maybe I should try to make friends with GIMP again, but it's such a pain to use.
|
Mouse crash
|
Topic: technology, opinion | Link here |
Nearly a year ago I bought a new mouse, a Logitech m705. I haven't been overly happy with it: of course, being modern, it doesn't have a middle button, and the side buttons don't fit well to my hand. I've assigned button 2 to one of them, and from time to time it vomits over my screen. The good news: it works.
Or at least, it worked. Today I accidentally ran out of desk while moving it, and it fell onto the carpeted floor, from a height of about 70 cm. That shouldn't be an issue, but it bent the right “button”, really a long strip of plastic, so that it fouled the left “button”. It's clearly moulded as a single piece, so there's not much opportunity for repair. Scraped some plastic off the side so that it didn't foul any more, but the switches didn't work properly, and the scroll wheel ran free. Clearly a write-off. And that a relatively expensive mouse from a reputable manufacturer.
But what do I replace it with? Wouldn't it be so nice if people built real 3-button mice again, with a couple of scroll wheels thrown in? But I can't find one. For the time being replaced it with the ALDI mouse that I used for teevee, the TV computer, in the process discovering that I had also disabled the keyboard (same receiver). But at least I can continue working.
|
More activity for the new dogs
|
Topic: animals, opinion | Link here |
We have two problems with Zagar, the elder of the two new dogs. Firstly, the name “Zagar” sounds almost indistinguishable from “Vago”, which is what Yvonne calls Zhivago. And secondly, he's getting more and more active.
We're still looking for an alternative name for Zagar. Currently we're thinking of Zolo, short for the kennel name (Zoloto), but it could change again.
Regarding the activity: that's not surprising, since he's only ever allowed out on a leash. So today, despite Yvonne's protests, we took him and Zhivago out into the paddock and let them run around together. That appeared to be just what they needed, and even Zhivago ran around like a wild thing. Clearly his operation has been a success. But it was so unexpected that I didn't even have a camera with me.
|
Tanya's paw still hurting
|
Topic: animals, opinion | Link here |
It's been two days now since Tanya hurt her paw, and she's still limping. Decided to take her to the vet, this time to Pene Kirk in Cape Clear, who diagnosed bruised knuckles and bandaged up the paw for a week. Also had her vaccinated while she was there.
| Tuesday, 7 January 2014 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Conservative revolution
|
Topic: general, opinion | Link here |
Cory Bernardi, a senator in the current Australian Federal Parliament, has written a book titled The Conservative Revolution. It was published a few months ago and appears to have received the attention it deserves—until some of the content was mentioned in the media yesterday. I don't know if anybody has read it, and the quotes didn't increase my interest, but Peter Jeremy pointed me at the Amazon review page, which is probably more worth reading than the book itself. I was amazed not to find any positive review, but finally I found this one, which was pounced upon mercilessly by other reviewers, and which now seems to have been removed. It's good to see that these things don't pass unquestioned.
|
No Exetel
|
Topic: technology, general | Link here |
For reasons I still don't understand, Exetel support asked me to make a PC (read: running Microsoft) available to them for remote maintenance today, so I installed their software on an old laptop and agreed to a call some time after 10:30. The call came really quite some time after 10:30, in fact at 18:30 as we were preparing dinner. So we had to postpone it until tomorrow. At least it gave me the chance to say to the engineer that the issue was not at my end. We'll see what happens tomorrow.
|
Ugly violent hardware
|
Topic: technology, opinion | Link here |
I'm in the process of choosing hardware for my next machine. It's not easy. Once upon a time you had a choice of two or three different processors, maybe different speeds, and any old motherboard. Now the choice of processor, motherboard and RAM is an order of magnitude more varied, and thus difficult. I've more or less settled on a Core i7 4771, my first Intel processor in over 20 years, and probably the Z87 chip set.
But in the process of investigating the products, I came across a really ugly trend: violent names. So far I have seen Vengeance, Fatal1ty, Sniper, Killer, DOMINATOR and Ripjaws. Does this really sell products? I suppose it's for the gamers, which doesn't say anything good about what kind of games people play, but I would be happy to pay a couple of dollars more to have a product without such obscene names. Ugh.
|
What happened to summer?
|
Topic: general, opinion | Link here |
The first week of January is past, but you wouldn't know it. The average temperature is fully 4.5° lower than average January temperatures in previous years?
It reminds me of Rudi Carell's 1975 song „Wann wird's mal wieder richtig Sommer?“:
Hopefully it won't stay this way.
| Wednesday, 8 January 2014 | Dereel | Images for 8 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Tidying up the garden
|
Topic: gardening | Link here |
We ultimately decided against the block of land in Snowgum Road, but Liam Crowley did us one favour: the potential seller, Craig Mayor, is a handyman and a gardener, and it looks like he'll be able to do quite a bit of work for us. He came along today to tidy up the garden, which I have almost completely neglected for the best part of a year. He's certainly efficient: he got more done in the day than both the Brooks Brothers did last May. Another couple of days and it could look good.
|
Exetel addresses performance issues
|
Topic: general, technology, opinion | Link here |
Exetel was due to call me this morning to address the network performance issues, so I first checked the status quo. That was interesting: 32.22 Mb/s downlink—and that on a 25 Mb/s connection. Had they fixed things? Tried testing via SkyMesh and got 33.62 Mb/s.
Is that possible? My first conclusion was that speedtest was broken. But of course this is an LTE connection, and it's capable of much more than that; it's just limited to 25 Mb/s, and if something goes wrong there, it could exceed the limit.
More to the point, though: file transfer. There, too, things looked better than before. Ten concurrent connections with axel showed speeds of 2.294 MB/s via SkyMesh and 2.393 MB/s via Exetel. That's almost acceptable. Have they fixed it?
Unfortunately, no. It didn't stay that way. This was early in the morning, and things
deteriorated during the day. This seems to be a clear indication of congestion. Exetel
support finally called me, at 16:56. The window was 10:30 to 17:00, so of course they
overran. And what did they want to do? A poor man's Microsoft equivalent
of axel. He tried to install µTorrent and BitTorrent on my Microsoft XP box, both
of which failed to start. Then he used a browser to run a concurrent download of files from
their mirror server—exactly the same test they asked me to do last week. Clearly this is their
standard test, though I had supplied them both those results and those on my speedtest page.
Still, the results looked good; scanning the download page the total probably exceeded 2 MB/s. But then I checked, and of course we were routing via SkyMesh. So I changed the routing, and the speed dropped to below 1 MB/s—exactly what I had been reporting. I pointed that out and showed the corresponding TRACERT output before and after the routing change:
|
||||||||||||||||||||||
I pointed that out to him, but he couldn't see it. Finally I showed him hop 3, via 1-208-181-180.cpe.skymesh.net.au. It wasn't until later that it occurred to me that he had maybe never seen a traceroute before. He said he understood, and that he would escalate the issue. They could have done that a long time ago. In the meantime, did another comparison with axel:
| Date | time | RSP | program | concurrency | server | speed (kB/s) | ||||||
| 8 January 2014 | 17:50 | Exetel | axel | 10 | debian.mirror.exetel.com.au | 1185 | ||||||
| 8 January 2014 | 17:56 | SkyMesh | axel | 10 | debian.mirror.exetel.com.au | 2718 | ||||||
|
More network interruptions
|
Topic: animals, general, technology | Link here |
Somehow I've had lots of trouble with the network connection to cvr2 lately. Only a couple of weeks ago I had to replace the network cable, but this morning it was off the net again. After a lot of searching, found this:
|
|||||||||||||||||||||||
Clearly Tanya has developed a taste for network cables. Surprisingly, I was able to find another cable and connect it to the switch in Yvonne's office—and it wouldn't work. Connected it to a switch in my office, and all was well. So it looks like a flaky switch as well as everything else. How happy I'll be in the new house with Ethernet jacks in every room.
| Thursday, 9 January 2014 | Dereel → Ballarat → Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Health check again
|
Topic: general | Link here |
Into town today to finally get the results of the routine blood tests I did nearly 2 months ago. In the meantime I've been so busy that I haven't found time for what I considered a routine matter. And I was right. Dr. Majid wasn't there, so Dr. Shadman (Ahmad) gave me the results; nothing out of the ordinary. I had intended to get my hair cut while I was in town, but I changed my mind when I saw the queue. So a trip into town just for that.
|
System upgrade pain
|
Topic: technology | Link here |
So it's time once again to upgrade my system. When it comes to installing FreeBSD, I wrote the book, but there's a vast difference between installing an operating system for the first time and migrating a large installation to a newer version. I've been working on this forever, and I thought that maybe the PKGng would make things easier. Maybe it will, too, but it won't make them easy. Spent most of the day, and at the end had some semblance of installation, but first I need to fix my scripts to install the packages that I don't have.
|
Tanya growth
|
Topic: animals | Link here |
Tanya continues to grow quickly. On 29 December 2013 she weighed 7.2 kg; now she weighs 9.4 kg, a 28% increase in weight in only 11 days.
|
Revisiting the building site
|
Topic: Stones Road house | Link here |
Back over to the Stones Road site this afternoon to put in more recognizable stakes for the house. They help, but it's still not useful for taking photos. Time for the soil tests to be finished so that we can put in our application for a planning permit.
| Friday, 10 January 2014 | Dereel | Images for 10 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More installation fun
|
Topic: technology, opinion | Link here |
On with my installation of FreeBSD 10 today, making a little progress. There are still rough edges and bits missing, though the result is still much better than things were 15 years ago, where I held installation workshops at conferences. But that's not enough. It's now lagging behind the competition. Hopefully Jordan Hubbard's return to the BSD fold will help there.
Just to add to the fun, installed the new copy of Microsoft “Windows” 7 that I received today. A completely different experience. And on the whole it went well, but then that's just a base installation. I only really use Microsoft to run photographic processing software, and most of the problems I have are the fault of that software, not of Microsoft. But the strange thing is how similar all versions of Microsoft are, to me at any rate; it's like having the same base system with a marginally tweaked window manager. Even “Windows” 8 has the same underlying structure (like hiding administrative things such as mounting remote file systems under “Computer” instead of “Control Panel”).
One thing that did change is a slight improvement in security: presumably as part of the installation process, the system detected the following suspicious behaviour on the part of the DxO Optics “Pro” installation:
|
|
||||||||
That's catching DxO in the act of writing away secret information about the length of my free trial. Looking further, beyond the end of the truncated message, shows a file with this content:
I have a license for DxO, but since this installation is something like a throwaway installation (I'll do the real installation on the old eureka once I upgrade that system), I went for the 31 day trial. And though this file contains a signature which will presumably stop you from modifying it, it's a clear candidate for deletion and being given a new 31 day trial. You'd think that they'd have thought of a cleverer way to do things.
|
Living on a gold mine?
|
Topic: Stones Road house, general, opinion | Link here |
How does the water flow round Dereel, in particular round our new building site? At first I thought it would run into Pinchgut Creek, which ultimately runs into the Barwon River, but the lie of the land makes that impossible. While looking, came across this site. It's very difficult to navigate, and I still haven't been able to work out if the creek that runs a few hundred metres from our house is Moonlight Creek or Misery Creek, though presumably it's more likely to be the former. But what I did find was the names of a number of mines, presumably gold mines. Unfortunately, apart from the name there's almost nothing about it. They haven't even realized that it's in Dereel.
On the first map, our property is on the junction of Stones Road and Grassy Gully Road, the T junction just left of the centre. It seems that all the mines are beyond the western boundary. But who knows? It would be interesting to understand a little about how to find gold.
| Saturday, 11 January 2014 | Dereel | Images for 11 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Tanya up to more tricks
|
Topic: animals | Link here |
Yvonne into the office this morning to tell me of Tanya's latest exploits: she had found some pills in foil and extracted them. Did she eat any? Yvonne thought there was only one pill left in the package, and that was still there. But who knows? Off she went again to Pene to have her stomach contents purged. Clearly we need to pay more attention to what she does, and also to where we store our medications.
|
Photo processing at a snail's pace
|
Topic: photography, technology, opinion | Link here |
House photo day again today. I've now migrated my Microsoft box dxo (“Windows” Vista 32 bit) to “Windows” 7 64 bit, and at Daniel O'Connor's suggestion I've renamed it dischord. Fired up the newly installed Olympus “Viewer” 3 and got a completely different view of the world. Suddenly I couldn't just select photos any more, I had to “import” them. Started doing that. I had 258 images, and after 5 minutes it had “imported” (copied across the network?) 20 of them.
I can't handle that kind of pain, and decided to go back to the old dxo disk. But the system had more pain in store: “Please do not power off or unplug your machine. Installing update of 1 of 18”. It took it another 15 minutes to power off.
Finally got “Viewer” running as well as I know how, and considered whether I should really run DxO Optics “Pro” on the output. Normally I output TIFF from “Viewer” and then use DxO to enhance the colour and brightness and then output JPEG. Decided against, and produced a direct JPEG output.
Finally got some output, and ran Hugin on the results. That worked, of course, but for reasons I still don't understand, enblend ran at a fraction of its normal speed, so slowly that I thought it had got stuck in an endless loop. Finally it finished, but it took over 15 times as long. Here ps output just before the end for this week and last week:
The image sizes were effectively the same (-f10572 specifies the width of the panorama in pixels). The program was the same (I rebuilt last week's panorama afterwards for comparison). What's causing it? I haven't had time to investigate, but it suggests that it's something in the JPEG output of “Viewer”.
|
Fixing the enblend port
|
Topic: technology, photography | Link here |
Some months ago a change to the FreeBSD port of vigra broke the build of enblend, which I maintain. When loading the final image, it produces voluminous “can't find” error messages, terminating in:
I must admit to being a little upset: this is not of my doing, and to fix it I need to look at the internals of vigra, which has given me pain in the past. Finally I've got round to looking at it, and it died in a completely different place, looking for a header file. So I reinstalled the vigra package from the packages repository, and was able to repeat the problem.
Clearly the problem is in vigra, so it made sense to build the port locally. Did that, installed it, and tried to build enblend. It worked! So somehow there's a difference between the vigra package and what I build locally. How I hate this stuff!
| Sunday, 12 January 2014 | Dereel | Images for 12 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Processing photos, continued
|
Topic: photography, technology, opinion | Link here |
On with my photo processing today. The verandah panorama took all morning, over 2 hours for the normal panorama and 3½ hours for the interactive flash version. What's the cause? Clearly it was worth finding out before spending days converting the remaining images, so processed the images the way I have been doing previously: first convert the image to TIFF with Olympus “Viewer” 3, and then use DxO Optics “Pro” to convert the TIFF to JPEG. Externally there wasn't much to be seen. Here's one of the 20 input images for the panorama; the others are similar:
=== grog@eureka (/dev/pts/7) ~/Photos/20140111 11 -> l C*/verandah-centre-0.jpeg
=== grog@eureka (/dev/pts/7) ~/Photos/20140111 12 -> identify C*/verandah-centre-0.jpeg
But something is significantly different. Processing the images gave me an amazing difference in time. Here the resource usage just before finishing:
Once again, that's a factor of 15 to 1 in the time. Memory usage was effectively the same, and so was the output size.
This isn't as simple as saying “broken JPEG”. Before being fed into enblend, the images are converted to TIFF, and there's nothing obvious about the TIFFs either. I'm baffled.
|
Specifying the new system
|
Topic: technology, opinion | Link here |
I've been investigating components for my new computer for a week now, and it's been tough. Finally I'm getting there, mainly by eliminating possibilities that seem too hard. The big issues remain the motherboard and the memory. An article in c't suggests that the Asrock motherboards are a good choice, and that's what Jürgen Lock recently chose. It works for him, so there seems little reason to compare the others.
And memory? There seem to be relatively few manufacturers to choose from: G.Skill, Corsair and Kingston. And almost all memory is dressed up for kiddies and has particularly violent names: Ripjaws, Sniper, Vengeance, Dominator. What happened to normal memory? On Jürgen's recommendation, decided to limit myself to Kingston memory, at least a company I recall.
As if to reinforce the necessity, once again I had problems with the switch in Yvonne's office. I now have 4 different switches with varying degrees of flakiness. Time to replace with a new switch.
|
Exetel througput problems: install new “Windows”!
|
Topic: technology, opinion | Link here |
So now I have demonstrated to Exetel beyond any reasonable doubt that they have a througput problem somewhere in their network. I've sent them comparisons with SkyMesh, they've tried their toy torrents personally, so there's really no doubt where the problem lies. So what's the next step?
To be fair, they're not blaming it on my operating system, but it's clear that they don't have the faintest idea. It's been ten days now, and I don't have the feeling that anybody has addressed the problem. I can't be bothered. Either they fix it, or I go elsewhere. I've given them until the end of the week.
|
Bassoon or buffoon?
|
Topic: music, opinion | Link here |
For some reason bassonists seem to be (even) more inclined to jokes than most musicians. Maybe it's the bizarre shape of the instrument, which led me to take these photos nearly half a century ago:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
For the record, I'm the one on the left in the first photo.
But it's not just my sense of humour: some years later, during rehearsals of Berlioz' Symphonie fantastique, where I was playing second, Roland Hill, the first bassoonist got up during the second movement («Un bal», during which the bassoons don't have anything to play) and waltzed around the back of the stage.
Back in the present, we're gradually watching the New Year's concert of the Wiener Philharmoniker in small portions. Today one of the pieces was „Schabernack“, by Josef Strauss. And what did we see?
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
There's a tradition of having things stick out the top of a bassoon; our attempts 50 years ago were very much in character. In this case, they had replaced the bell with giant bassoon reeds. The first bassoon even had the same somewhat unusual colour as the reed he was playing. It seems to have taken the camera people by surprise. Those are the only shots they included, and the first one doesn't even show the second bassoon, who had also changed his bell. In general it seems that they preferred to take views of specific players, such as this one, who I suspect is related to one of the cameramen:
|
|||||||||||||||||||||||
| Monday, 13 January 2014 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Powercor: so nice, so nice, we do it twice
|
Topic: general | Link here |
Short power failure this morning at 8:45, while I was cleaning my teeth. I didn't notice until I wondered why the radio had stopped.
Out of habit, didn't reset the alarm clock until the evening. It didn't help much: at 23:14 there was another failure.
|
Bluetooth audio
|
Topic: technology, multimedia, opinion | Link here |
I've been listening to Radio Swiss Classic via the web for some time now. That meant moving the Android tablet to the fridge, where the mini-Hi-Fi system is located, and connecting it with a cable. It also meant that I couldn't do much else with the tablet while it was playing music.
Clearly a case for a Bluetooth audio adapter. Do they exist? Yes, and they cost next to nothing, $6.50 including postage. I ordered one from eBay last week, and today it arrived. There's almost nothing to it: a USB connector at one end, for power, an audio jack at the other end, and very little in between:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
It even comes with a far-too-short cable, but no instructions. Strangely, connecting it up Just Worked, and now I can even put the tablet in the next room, further than the promised 5 metre range.
|
Fire in Grassy Gully Road!
|
Topic: general | Link here |
On IRC today, Chris Bahlo told me that there was a bushfire in Grassy Gully Road. Sure enough, it showed up in the Bushfire Overview.
Grassy Gully Road ends (west end, where Google Maps has placed the marker) almost directly outside our new property. Where could a fire happen there? And what would it do? Off to the site to find a CFA truck at the end, talking to a car going in the other direction. Before I could stop and get out, he drove off. He was the only sign of anything fire-related. Slowly drove the length (900 metres) of the road, looking carefully to either side, but couldn't see anything remotely resembling the remains of a fire. Got back home and discovered that it was a false alarm.
| Tuesday, 14 January 2014 | Dereel | Images for 14 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Summer comes with a vengeance
|
Topic: general, opinion | Link here |
Last week I was complaining about the cold. Today the weather made up for the slow start to January: we recorded the highest temperature ever, 45.6°. The maximum outside temperature in the first week of the month was 25.3°; the minimum I recorded today was nearly 5° higher:
And that has been enough to bring the averages back in line:
And they're planning another 4 days of this weather! Why can't it just be normal?
But was it really the highest temperature? Five years ago I measured an estimated 46° on the verandah:
|
|
|||||||||
It's really difficult to compare: today's highest temperature in that location was 42.1°, but things have changed in those five years. At the time there was almost no vegetation round the verandah, and now there's lots:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
USB says: time for a new system
|
Topic: technology, opinion | Link here |
Into the office this morning to find the keyboard and mouse dead again. Disconnecting and reconnecting didn't help:
Finally, after removing everything and reconnecting it, I got things to work. Almost. My weather station software still wouldn't run, and more investigation showed that the USB subsystem wasn't finding it. Defective station? I have a second one, so connected that. No go. Finally connected to Yvonne's machine, and it recognized it without a problem. So recompiled the software for her machine (i386, not amd64), and all was well.
For about an hour. Then the station stopped reporting again. This time it wasn't related to the hardware: today was the hottest day since I wrote the software, and the station started reporting 0% relative humidity, which I had considered to be invalid. A minor tweak and it worked again—until Yvonne shut down her machine in the evening, of course. Until I fix things, I'll have to leave her machine on all the time.
As if out of sympathy, my new Bluetooth dongle stopped associating. How I love Android: it's too leet to report errors. Tell it to associate with a Bluetooth device and it will go away, say “connecting”, and then stop. If it says “Connected”, it's connected. If it says nothing, it's not connected.
It's not clear why the connection broke, but possibly because the dongle gave up when the tablet went out of range. Tried to remove it from the power supply, but ended up with only the case, which came off the device when I pulled. But I finally got it back together, having gained some experience in the process.
But clearly all this (and more—I've had several X crashes lately) is telling me I need a new machine. Finally finalized the configuration that I've been working on for over a week now, and placed an order. Power supply and memory currently not available, no indication when it will be. Looks like it'll be a while yet.
|
Exetel: on the way out?
|
Topic: technology, opinion | Link here |
Exetel have still done nothing to fix my network throughput problems. Though they appear friendly and cooperative, they also appear to have no understanding of the issues or how to address them. On Sunday I gave them until the end of the week to fix the problem. So far they have done nothing. So I called up sales to tell them that I was about to terminate my contract because of breach of contract. Spoke to Paula, who didn't seem to understand the issue and told me I'd have to pay $100 early termination fee. Finally she suggested that it was a matter for support, and promised to connect me with a support manager. She put me through to Aisha, a support engineer, who finally connected me to Theje, a manager.
Theje told me that no, in fact they had been doing lots of work, including a new server (huh?), and that the network supplier had problems and was addressing the issue. The supplier is NBN, of course, and their link works fine for other RSPs. He wanted another session of an engineer connecting to my system, and also to install Yet Another Software Package on my Microsoft box, both of which I refused because it was a waste of everybody's time. I explained that I had done everything they have done before I even put in the ticket, that I was using axel to set up 10 concurrent TCP connections, and that the results different greatly between Exetel and SkyMesh.
He put me on hold for a while and talked to somebody—a real engineer, maybe—and came back and said ”if I understand it correctly, you're doing single-threaded transfers”. Clearly he didn't understand. He also then claimed that there was no difference in speed between Exetel and SkyMesh. And they had proven conclusively that the matter had nothing to do with congestion. Finally he conceded that they should try to re-route the connection, and then he would get an engineer to run a test on my system. I refused and said that they should contact me and I would perform the tests myself. That's supposed to happen tomorrow. I won't hold my breath.
|
NBN: An NSA plot
|
Topic: technology, opinion | Link here |
For some reason Edwin Groothuis wanted to know about physical access to the Radiation Tower tower today, in particular whether there was a sign on the gate. No:
|
|||||||||||||||||||||||
But he meant the access to the immediate compound, and that's not accessible to the public. I had, however, taken a photo of the entrance to the Enfield tower last year:
|
|||||||||||||||||||||||
|
|
|||||||||
NSA? Yes, of course we knew the name 15 months ago. But since Edward Snowden everything appears in a very different light. Is this some sinister plot?
| Wednesday, 15 January 2014 | Dereel | Images for 15 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More hot weather
|
Topic: general | Link here |
Today wasn't quite as bad as yesterday, but it was still one of the hottest on record. Once again spent nearly all day indoors.
|
Another week, another dog's name
|
Topic: animals, opinion | Link here |
What's in a name? In the case of a dog's name, it should be something that the dog will recognize and other dogs won't. When we bought our new dogs, the older one already had a name: Zagar. But it soon became apparent that the name sounded too close to “Vago”, which is what Yvonne calls Zhivago. So we changed it to “Zolo”, for no better reason than that it's the kennel name. But now we're not happy with that either, so now we've changed the name conceptually to Nikolai (Николай), though we'll call him Nikolai with a long ɪː. Hopefully it'll stay that way.
|
Another power failure
|
Topic: general | Link here |
Yet another power failure this afternoon at 16:25, reminding me that it's time for a new UPS in the lounge room. We should find a better solution for the new house.
| Thursday, 16 January 2014 | Dereel | Images for 16 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Heat wave continues
|
Topic: general | Link here |
For once the forecast of the Bureau of Meteorology has proven correct. Today was the third day of a prolonged heat wave:
And another day to go! I hope they're right that it'll be over then.
|
Tanya's weight
|
Topic: animals | Link here |
One of the reasons Tanya is gaining weight so much appears to be that she's getting too much to eat. Her ribs are only barely evident, so we're cutting back. It hardly shows: today she weighed 10.8 kg, up from 9.4 kg 6 days ago.
|
Still more network problems
|
Topic: animals, general, technology | Link here |
I had put an Ethernet cable over the floor last week after Tanya had chewed through the other one. Somehow I wasn't expecting her to do it again, and to a certain extent I was right:
|
|||||||||||||||||||||||
Despite the appearance, it still works. But clearly we'll have to find an alternative. For the time being I have cut the old cable into 2 m lengths and spread them round the house; if she (or Nikolai) starts chewing on them, we're more likely to catch them and make it clear that it's not allowed.
|
Olympus autofocus
|
Topic: photography, opinion | Link here |
It still seems that autofocus is the biggest issue in photography today. The Olympus OM-D E-M1 has raised the bar in that respect, but only with Micro Four Thirds lenses. With Four Thirds lenses things aren't so good.
I took the first photo of the cable that Tanya chewed with the Zuiko Digital ED 12-60mm F2.8-4.0 SWD, a Four Thirds lens. Under the lighting conditions, it was barely acceptable. But I needed to come closer than the 60 mm focal length, so I tried it with the Zuiko Digital ED 70-300mm F4.0-5.6 lens. And that didn't work at all. I had to get out the old E-30 and use that, and it wasn't very happy either. What I finally ended up had the focus in the wrong place, so basically it's useless (second image):
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
I suppose it's about time to apply the latest firmware upgrade to the E-M1; it claims, without further details, to improve autofocus.
| Friday, 17 January 2014 | Dereel | Images for 17 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Breaking temperature records
|
Topic: general | Link here |
One of the temperature records that I haven't put on the web yet is a record of the temperature in the greenhouse. I have a minimum/maximum thermometer in there, originally designed for domestic use as inside/outside. I have the “inside” part mounted at the top of the greenhouse, and the “outside” sensor is at the bottom in the shade, and I keep the records in a (paper) notebook.
Clearly temperatures in greenhouses are higher than outside when the sun is shining, and I have opened all orifices and covered the greenhouse in shade cloth. Nevertheless, today I recorded the highest ever temperature: 50.4°!
That wasn't the only record. Today was the promised last day of the heat wave, and the Bureau of Meteorology kept its word, but not before the outside temperature hit 46.6°. Here's an overview of the entire heat wave:
|
Reception problems: insights?
|
Topic: general, technology, multimedia, opinion | Link here |
I've been puzzling about TV reception quality for years now, without coming to any obvious conclusions. Radio has also been very variable as well, and from time to time I've wondered if the issues are related. But today I had a surprise: just out of the shower, there was a power surge or similar, not enough to cause any devices to power cycle, but enough to make at least one UPS scream. And the radio reception went to hell! All I could hear was noise. So: is this a (or more than one) rogue UPS? Chasing that one down could be fun.
|
Tracking down the Exetel problems
|
Topic: technology | Link here |
Call from Theje of Exetel today to address the ongoing network problems. He claimed that they had changed the routing. I had expected something like this, and had already saved a traceroute output of what I currently had:
=== root@eureka (/dev/pts/5) /home/grog 5 -> traceroute cdn-edge1.mel4.internode.on.net
So it was relatively easy to compare the new routing:
=== grog@eureka (/dev/pts/14) ~/Photos/20140114 4 -> traceroute cdn-edge1.mel4.internode.on.net
Pointed this out, and after a bit of ”hold on, need to discuss” they patched me into a conference call with a Real Engineer, who rather smugly suggested that I was mistaken in saying that they had not changed the routing. But it seems that previously I could have had one of many routes, whereas now I was hard-wired to this one. Not surprisingly, nothing much changed. Spent over an hour discussing the issues, including my measurements two weeks ago, which at least the Engineer (name unknown) understood. He tried various things, including gaining direct access to pain (my Microsoft XP box) bypassing the router; like me, he suspected that it might be involved. Unlike me, he didn't believe my conclusions after testing.
The tests showed that no, direct connection made no difference, and yes, indeed, changing the routing has a noticeable effect on throughput—in almost every case, it got worse:
| Date | Time | RSP | Server | Ping | Downlink | Uplink | Comments | |||||||||
| 17 January 2014 | 11:42 | Exetel | Internode Melbourne | 39 | 11.39 | 4.13 | Link | |||||||||
| 17 January 2014 | 11:44 | SkyMesh | Internode Melbourne | 53 | 20.02 | 3.67 | Link | |||||||||
| 17 January 2014 | 12:18 | Exetel | Optus Melbourne | 54 | 7.92 | 3.86 | Link | Directly connected to pain.lemis.com | ||||||||
| 17 January 2014 | 12:19 | Exetel | Internode Melbourne | 31 | 12.82 | 4.03 | Link | Directly connected to pain.lemis.com | ||||||||
| 17 January 2014 | 12:20 | Exetel | Optus Sydney | 52 | 7.63 | 3.71 | Link | Directly connected to pain.lemis.com | ||||||||
| 17 January 2014 | 12:22 | Exetel | Internode Sydney | 47 | 9.83 | 3.52 | Link | Directly connected to pain.lemis.com | ||||||||
| 17 January 2014 | 12:42 | Exetel | Internode Melbourne | 40 | 6.49 | 3.51 | Link | Changed routing | ||||||||
| 17 January 2014 | 12:45 | Exetel | Internode Melbourne | 35 | 11.15 | 4.14 | Link | Changed routing | ||||||||
| 17 January 2014 | 12:46 | Exetel | Internode Melbourne | 45 | 2.59 | 3.47 | Link | Changed routing | ||||||||
Then he got me to install Yet Another Program on pain, flashget, an irritating program that grabs marked texts and interprets them as URLs, which it downloads concurrently, which bleeds through other windows and which won't go away. At least it works. It showed the same results as all the others: the link seems to max out at a not-very-stable 1 MB/s.
Then later in the afternoon I received an email:
That's enough! I've spent hours working with these people, and today I thought they had finally come up with somebody who understood the issues. Wrote an appropriately worded reply and posted a corresponding comment on Whingepool. That got a reaction: a company representative with the appropriate nick shaming asked me to send him details of the matter. And a few hours later, before I had even noticed the posting from shaming, I got another email:
Coincidence? In any case, if they're going to do something about it now, I'll be happy enough. And despite all the problems, I've found the people polite and solicitous, unlike, say, Telstra.
|
Domestic network woes
|
Topic: technology, general, photography, opinion | Link here |
There's something funny in my domestic network infrastructure as well. I've reported various problems which seem to be related to the switch in Yvonne's office, and so far the only one I have identified beyond reasonable doubt are the two cables that Tanya chewed through.
Today I tried to read in some photos from my camera. I do that on lagoon, Yvonne's machine because of the USB issues on my own (current) machine. And they get copied via NFS to my machine. The network topology involves a cable to a switch in the cupboard in Yvonne's office, then a cable under the house to my office, and thence into the main switch. This is the same switch that has another under-house cable to the cupboard where I keep cvr2, the video recorder machine, which has also had network problems.
Today I couldn't copy the files:
Tried changing switches again—I have at least 5, most of which seem to work. But nothing. MTU? Hardly an issue if it has been working for years with no problems. I can see at least 3 potential problems:
Switch. This is still the most likely, but why have so many switches failed? In any case, next week I'll get a new one, so I can rule out that one.
Cables? I terminated the under-house cables myself, and it's quite on the cards that they were once marginal and now decidedly flaky. I'm not going to try to fix that one; the next house will have professionally installed Ethernet in every room.
Power issues, such as ground loops? Yvonne's office is on a different power circuit from mine. But so is cvr2.
Of course, there's no reason to assume that there's only one cause. The next thing to do is to order some new Ethernet cables, which I'll use if the ones under the house prove to be the problem.
| Saturday, 18 January 2014 | Dereel | Images for 18 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Time for better HDR software?
|
Topic: photography, technology, opinion | Link here |
House photo day again today, and again I've tried making HDR images of all of them. And again things were less than perfect, with lots of ghosting:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
How do I fix that? Other software seems to be about the only way. Currently I'm using align_image_stack (part of Hugin) and enblend to merge the images. Here's the core of my HDR script:
That works well for static images, but it has no provision at all for avoiding ghosting. What are the alternatives? For some time I've been meaning to read the article The Grumpy Editor's guide to HDR with Linux, and today I finally got around to looking at it. It's not a bad article, but it's 7 years old, and it doesn't even mention ghosting. And all the images carefully avoided the problem.
So went off looking for packages on Google, and predictably found plenty. Are they any good? The only way to tell is to try them out. And that involves lots of work. Watched a video about HDR Expose 3, which showed some interesting stuff, but also far too much mouse activity to find image sets. All sets were shown graphically on a window, one line per set, so at any one time you can see 4 or 5 sets. Today I took 78 sets of images. How do I find them with such an interface? It's possible, of course, that they've taken that into account for an “advanced” user, but I wouldn't hold my breath.
|
Goodbye Jason Gale
|
Topic: general, opinion | Link here |
Sad news via Facebook today. Jason Gale, son of my cousin Jan, died suddenly on his honeymoon a couple of days ago. That's sad for many reasons. I was never close to him—having spent most of my life either overseas or holed up in the country, I don't see any of my family much, and he grew up while I was in Europe—but it's really sad that the main medium of communication of such events is by Facebook. Somehow that's far too public.
| Sunday, 19 January 2014 | Dereel | Images for 19 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Garden flowers in mid-summer
|
Topic: gardening | Link here |
I should have taken my monthly garden flower photos 4 days ago, on 15 January 2014, but the temperatures were so high that I couldn't face it. Finally got around to doing it today; things have suffered considerably. Earlier in the week I had topped up the pond to its normal level, the top of the fountain. Now it had dropped by fully 15 cm:
|
|||||||||||||||||||||||
As a result, there's wasn't much to see. Despite watering, even some relatively drought-tolerant plants suffered badly:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Many flowers, such as Fuchsia and Abutilon, stopped flowering altogether. Surprisingly, a Cyclamen on the verandah is still (just) flowering:
|
|||||||||||||||||||||||
And the potted Hibiscus, now on the north terrace, looks surprisingly healthy:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
Relocate the new house?
|
Topic: Stones Road house, opinion | Link here |
The soil tests for our new house still haven't been done. They should have been done 2 weeks ago at the latest. That's bad, because they're on the critical path: we can't apply for a planning permit until we have the results. On the other hand, it gives us the ability to reconsider the location of the house.
The big issue is the pond in the eastern part of the property. We had originally intended the house to be just to the east, and to include the pond in the planning, but the more I think of it, the more it seems to be in the wrong place. Today over to walk around and consider whether to put the house to the west of the pond, which is under the trees in the mid-foreground, just to the right of centre of this photo.
|
|||||||||||||||||||||||
The photo faces east, so the house would be just in front of those trees. We need time to think about it.
| Monday, 20 January 2014 | Dereel | Images for 20 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
New house: cast in concrete
|
Topic: Stones Road house | Link here |
Called up Tom Tyler this morning to discuss the soil tests and the location of the house. No, the soil tests haven't been done yet, but the site survey has:
|
||||||||||||||||||||||
And that doesn't cover the area to which we were thinking of relocating, to the west of the trees. So we have the choice of having the survey done again—certainly a possibility—or staying where we originally intended. The real issue is the pond, which is in the wrong place. But we could always move it if it proves to be a problem; the soil from excavating the building site would probably fill it. So we'll leave the house where we planned.
|
Playing with DraftSight
|
Topic: Stones Road house, technology, opinion | Link here |
Tom Tyler's email included an attachment with the name Not used for HTML emails.dwg. Apart from the silly “name”, what's that? Jürgen Lock came out with the discovery that it's an AutoCAD file format, and Jashank Jeremy was able to decode it: it's the site plan above. He also came up with the information that there's a free compatible CAD package called DraftSight that can process them. Installed that and took a bit of a look: yes, it seems to be relatively powerful. Now I just need the time to learn how to use it.
|
Ashampoo Photo Commander
|
Topic: photography, technology, opinion | Link here |
Last week I was considering buying Ashampoo Photo Commander 11, and gradually came to the conclusion that it wasn't worth it. But I bought it anyway. They had a special offer, and along with Movie Studio (explicitly not Pro) it cost me $20. Was it worth it? One problem is clear: in the documentation they claim to have exceptional customer service, but so far neither of my questions about the products have been answered.
One thing that it can definitely do is optimize photos. I've been using an old version of their “Optimizer” program for some time, but it occasionally goes crazy and messes up photos, and it can't handle the size of images I've been generating lately. “Commander” also has an optimization function, and so far it doesn't seem to have either of these limitations.
But the interface! Like all Microsoft-space programs, it has sticky cwds, and it requires much more mouse pushing. With the old version you clicked on “optimize all”, answered the somewhat superfluous message “Are you sure? If so, you can recover”, and it did its thing. Now you need to select all images (^A, once it has finished trying to display them), press Alt-1, wait again, select Next, Choose Effects, select Full Optimization, OK and Start. Surely I'm not the only person who thinks there should be an easier way.
On the positive side, it seems to recover an amazing amount of detail. On Saturday evening I took a number of photos of Chris Bahlo and the animals. On one occasion, the flash didn't go off, but the camera didn't notice, so I got a completely black image:
|
|||||||||||||||||||||||
I forgot to remove it before putting it through the optimizer—and it actually recovered some detail:
|
|||||||||||||||||||||||
And for the rest, it comes close to my desire for automatic exposure compensation. So, I suppose, it was worth the relatively low purchase price.
|
Ports pain all over again
|
Topic: technology, opinion | Link here |
The hardware for my new machine should arrive some time this week, so it's high time to have a system to put on it. I've already installed those packages that I could find in the FreeBSD repository, but that was a while back. PKGng has an upgrade command: type pkg upgrade and it should upgrade all installed packages.
That's the theory. What I got was:
Further investigation showed that xorg-server had failed to build on the build machine and had thus been removed from the repository. Hopefully that sort of thing won't happen again.
| Tuesday, 21 January 2014 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Coursera in the age of NBN
|
Topic: technology, general, opinion | Link here |
A few months ago I participated in a few courses from Coursera. Getting the videos was difficult, and it took me several hours each week to download them.
They were taking up too much time, so I stopped for a while, but now there are a couple that are quite interesting: Programming Mobile Applications for Android Handheld Systems and Human Evolution: Past and Future. The second will presumably not take up too much time, and the first sounds interesting enough to expend some time on it if necessary.
Today both courses started, and I loaded the videos. What a difference! All were done within a couple of minutes. It's times like this that remind you what an improvement the National Broadband Network is.
And the courses? Evolution is interesting. Android looks like it might take me into an area I'd rather avoid. But we'll see once I have the software up and running.
|
Preparing for the new machine
|
Topic: technology | Link here |
Another phone call to Umart today about my new machine. They promised to find an ETA for the still-missing components and get back to me with details. They were about the best I could expect: things have been shipped.
In preparation, started building those ports that wouldn't install from the PKGng repository. First was just configuring the packages, which took an hour of pressing Return. How many packages did I “configure”? Hard to say, since all that information is stored in a database that doesn't easily (to me, anyway) show what packages it belongs to. But it must have been several hundred. And then I went through the first build target (ports-try, which tries to build every port and continues if it can't). Chrom* took all day to build! I wonder if I'll be finished by the time the hardware arrives.
| Wednesday, 22 January 2014 | Dereel | Images for 22 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
FreeBSD 10.0: Not ready for prime time
|
Topic: technology, opinion | Link here |
Once again I've spent some considerable time installing the latest and greatest version of FreeBSD on a new disk in preparation for the new machine that is on its way. Coincidentally the FreeBSD project announced 10.0-RELEASE today, though I have been tracking it for some time.
Yesterday I started a ports-try target, which tries to install every port in my list of ports. It doesn't have to succeed: if not, it just goes on and builds the next one. Then I can come back and look at the ports that fail to build.
Why should a port fail to build? Hasn't it been tested? Well, yes, but there are over 20,000 ports, some of which are of only marginal use. But it seems that even a core group of ports doesn't necessarily build. I saw that on Monday with X. And today many ports failed. racket failed with errors that didn't make much sense to me:
That looks pretty basic to me. It's maintained by Jimmy Olgeni, and I don't need it urgently, so sent him off a message. Hopefully it'll get fixed.
Next came GIMP, which failed in some dependency. But avidemux2 took the cake:
Redefine free() as an error? At least it gave me the location:
So clearly this is a build configuration problem: ADM_LEGACY_PROGGY should be defined. You'd think that was straightforward enough, but this thing uses cmake, which I've never learnt, at least because the brightly coloured output irritates me. But it suggests that the problem is relatively easy to fix.
Things went on. Jürgen Lock, always ready to help, tried building GIMP and succeeded. That's one point each for him and me: he can build GIMP, I can build enblend.
This is a catastrophe! This is a freshly released system. The reason I have to build these ports in the first place is because they're not in the package repository, and that means it doesn't build on the build machines either. To paraphrase a saying, “It doesn't build. Ship it!”.
How could we get here? 20 years ago FreeBSD was an example of good, easy software building. To build the base system, you entered cd /usr/src; make world. To build, say, GIMP, you entered cd /usr/ports/graphics/gimp; make install. Now just about any other system outruns us.
To be fair, PKGng is still very new. But it seems that one of the big problems was the decision to replace gcc with clang. Why did we do it? For political reasons. gcc is now under GPLv3, that horrible enemy of free software!
That kind of politics makes me sick. It seems to base on a misinterpretation of the license, and for most people it doesn't make the slightest difference. It seems that it does for at least some of our corporate sponsors. But that's no reason to make the system unusable.
I can't put up with this. Downloaded a 9.2-RELEASE CD and started again. That wasn't plain sailing either—9.2 doesn't really like PKGng very much:
That proved to be because the base system installed a (now) invalid file /usr/local/etc/pkg.conf. But even after removing it (the current version is located in /etc/pkg/FreeBSD.conf), things didn't work:
Nothing got installed, so back to building everything. At least that seems to work, though I'll see for sure at the end.
I find this incredibly depressing. It's almost enough to make me change platform. But there's no other that is (still) as easy to use.
|
Grass tree in bloom
|
Topic: gardening | Link here |
Years ago in Wantadilla we planted some seeds—1000 in total, I believe. Most of them didn't survive, but one that did was a Xanthorrhoea quadrangulata, also known as a Mount Lofty Grass Tree. Over four years ago we planted it in the northern (“Japanese”) garden, where it didn't look up to much. It's the thin spike in front of the rocks in this image:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
And now—finally!—it's producing a flower spike:
|
|||||||||||||||||||||||
I wonder how much bigger it will get.
|
Raw images: Not with Ashampoo Commander
|
Topic: photography, technology, opinion | Link here |
Since getting my new Olympus OM-D E-M1 my photo processing hasn't got any easier. I used to use DxO Optics “Pro” to convert the raw images and enhance them a little, then pass the images through Ashampoo photo optimizer. Now DxO no longer understands my combination of camera and lens, so I have to use Olympus “Viewer” to convert the images to JPEG. And now I have the new Photo Commander 11 to do the optimization for me.
But “Commander” can also process raw images. In the cases where distortion correction isn't important, why not just use it as a one-stop conversion program? Tried that today and got my answer:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Run the cursor over the images to compare with the partner. In each case the first image is produced by my current method, and the second was committed by “Commander”. Clearly it's useless for raw images. Looking at the divergent dimensions (4640x3472 pixels instead of 4608x3456), my guess is that it's a poor adaptation of dcraw. That's no apology for completely breaking the EXIF data, of course:
| Thursday, 23 January 2014 | Dereel → Cape Clear → Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Another power failure
|
Topic: general | Link here |
Another power failure this morning at 2:10.
|
Vaccinating Nikolai
|
Topic: animals | Link here |
Yvonne is planning to attend a puppy training course next month, and of course one of the things they want is a vaccination certificate. That's fine for Tanya, but we don't have Nikolai's certificate: when we bought him, Ron couldn't find it, and he's currently not able to attend to the matter. So off today to Pene Kirk in Cape Clear to have another vaccination done.
|
Bore: no easier than last time
|
Topic: Stones Road house, opinion | Link here |
In the afternoon, to the new building site to meet Des Murray about having a bore done. I had hoped that, given the lie of the land, water would not be far below. That appears not to be the case, unfortunately: Des thinks it could be as much as 80 m below the surface. On reflection, that makes sense: low-lying areas tend to fill up with sediment, so the real rock surface is deeper than it would be on a hillside. Still a nuisance—we could be in for over $10,000 for the whole thing.
|
Exetel problem, next chapter
|
Topic: technology, opinion | Link here |
Call from Exetel support today: they wanted access to a BSD machine. Why? They couldn't tell me. But after some discussion on IRC, where in particular Edwin Groothuis was very much in favour, set up an account on stable-amd64, the machine that I'm still in the process of building.
How do you give access to somebody from outside when connected via NAT? The simplest solution, it seems, was not to do so. Instead set up PPPoE on the box and connected directly.
That worked relatively well, and the tech (Glen) basically repeated Yet Again all the things that I had done before. Performance was even worse than usual. Played around with axel and got speeds of only round 1 MB/s. Then Glen tried 3 concurrent axels. What good should that do? I've already established that the throughput trails off badly with concurrencies of more than about 5, and these tests were done with 10. But we tried anyway, and how about that! The transfer rates were 829 kB/s, 826 kB/s and 744 kB/s, a sum of 2.399 MB/s. That's almost acceptable.
But why? We had a total of 30 concurrent TCP sessions. Why should that improve things? It's interesting to note the ping times. Here the summaries, first during the transfer, and secondly without other traffic:
In other words, there is no noticeable difference. That in itself is strange. But it's reasonable to assume a latency of 50 ms. To fully load the network with one connection and that latency would require a window of 25 × 0.05 bits, 1.25 Mb or 156 kB. That should be trivial. With 30 connections the requirement would drop to only 5.2 kB. Surely it's not that bad.
But that's the issue: we're 3 weeks into this problem, and so far nobody apart from myself has done any packet tracing. And they seem to think that if they can saturate the connection with 30 TCP sessions, the problem is solved. As I reminded them, from the end user point of view we're seeing double the performance from SkyMesh that we're seeing from Exetel. Hopefully they'll finally do some real trouble-shooting.
|
Strange weather patterns
|
Topic: general | Link here |
The heat wave of last week is over, and the temperatures were pretty constant for most of the week—until today. Then, suddenly, we had high temperatures again:
It's fascinating how temperatures can vary by that much from day to day.
|
New computer arrives—almost
|
Topic: technology | Link here |
Phone call from TNT this afternoon: they had attempted in vain to deliver my new computer, because the driver couldn't find the address in his GPS navigator. That's understandable—I've ranted enough about the poor quality of the maps round here—but why didn't he call me?
In any case, that proved not to be a problem, because Yvonne was just about to go into town, so she picked it up. But it's still missing memory, SSD and power supply. Maybe, if we're lucky, that will arrive tomorrow.
| Friday, 24 January 2014 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Tracing the Exetel connection
|
Topic: technology, opinion | Link here |
It's now been over three weeks since I submitted my ticket to Exetel. Yesterday's “testing” gave us some new information, but they still don't seem to have done the obvious thing and trace the connection.
Instead a got a call from Exetel (Eroshan Jayaweera, if I got the name right) asking questions that I had answered several time in my speed test page. And then they had an astounding request regarding these results of the tests I made yesterday morning:
| Date | time | RSP | program | concurrency | server | speed (kB/s) | Comments | |||||||
| 24 January 2014 | 9:10 | Exetel | axel | 10 | octopus.com.au | 1077 | ||||||||
| 24 January 2014 | 9:10 | Skymesh | axel | 10 | octopus.com.au | 2122 | ||||||||
| 24 January 2014 | 9:13 | Skymesh | axel | 30 | octopus.com.au | 2476 | ||||||||
| 24 January 2014 | 9:14 | Exetel | axel | 30 | octopus.com.au | 1427 | ||||||||
They wanted screen shots of the tests. What kind of confusion could have given rise to that idea? Eroshan apologized and said that it was for management. Of course, if Exetel management is that confused, it might explain something. Rather than fight, I created a new page with the information. I wonder how anybody with so little understanding will be able to interpret it, when even for experienced people it says as good as nothing.
And still no traces! I've done some traces before, but they were of axel transfers, and analysing that many TCP connections is like pulling teeth. The real question is why a single TCP connection can't saturate the link, so I tried transferring a 10 MB file from their mirror server. The results were interesting. The first trace showed a receive window (my end) of 65535 and a send window of only 30080. Further investigation showed that I hadn't configured RFC 1323) extensions. To quote the entry I put in /etc/sysctl.conf:
After that I got a receive window size of 3217152, about a third of the total transfer size, but the send window remained at 30080, and the transfer rate was no different:
=== root@stable-amd64 (/dev/pts/10) /home/exetel 33 -> ftp http://debian.mirror.exetel.com.au/debian-backports/pool/main/a/amarok/amarok-common_2.4.1-1~bpo60+1_all.deb
That's 4.5 Mb/s, or 18% of the bandwidth. Looking at the trace (from frame 290) showed lots of:
This is done with my own hack of tcpdump to show easily readable incremental times between the individual packets in the first column, here in microseconds. I've edited the host names to make the output fit on the web page. The National Broadband Network link is an Ethernet bridge, so raw packets go over the link.
But there's a problem here: the excerpt above shows that 4 packets (5936 bytes) arrived in 124 µs. That's not even the complete truth: it doesn't include the PPPoE overhead. Based on traces on the SkyMesh link, which doesn't have this overhead, the MTU is really 1520. So in 124 µs I received 6080 bytes. That's 49 Mb/s.
How can that be? This is a 25 Mb/s connection. Well, no. It runs LTE, which can easily do 100 Mb/s on a good connection. NBN has to limit the data rate. But how? I've already seen Speedtest results where at least briefly the indicated speeds exceeded 50 Mb/s. This trace is another indication that their rate limiting might be less than perfect.
But that's obviously only half the issue. At that speed, the file would have been transferred in 1.6 seconds. In fact, it took more than 10 times as long. And it's not hard to find the reason. In fact, it's directly before the previous example, starting at frame 284:
Over 50 ms delay! How could that happen? Looking only from this end, it looks as if it's not a window issue: I've been acking religiously, and even if a single ACK got delayed, it wouldn't slow things down that much. And that's the real problem. Here's a statistical overview of the trace information:
| Total packets transmitted: | 10761 | |
| Total time (from trace): | 17.637747 s | |
| Number of packets delayed by more than 10 ms: | 612 | |
| Elapsed time for these packets: | 12.736507 s | |
| Number of packets delayed by more than 30 ms: | 64 | |
| Elapsed time for these packets: | 2.444560 s | |
With a well-functioning TCP connection the packets should come in at roughly the speed of the connection, in this case 25 Mb/s. So a packet with an MTU of 1520 would take about 49 µs to arrive, and that's what the trace shows most of the time. But roughly 6% of the packets are delayed far beyond this level, and they account for 72% of the time. Without them the transfer would be nearly 4 times faster.
The proximity of very good and very bad performance raises an obvious question: is there a connection? Looking through the trace, there's nothing obvious. But then, this kind of trace isn't obvious. What I need are better tools.
Another issue with this trace is that it shows a number of dropped packets, the first being recognized at frame 307. This dropped packet took 62.5 ms to resolve, but during this time 18564 bytes were transferred, a rate of 297 kB/s. Clearly it would be good to understand why the packets are dropped, but that alone won't explain the performance.
This is only the view from one side, of course, but it's interesting because both of the delays and the overly fast arrival of the data. Does NBN have some involvement after all? Clearly what needs to be done is to trace simultaneously from the other end. I wonder if we can finally get Exetel to do some useful measurements.
| Saturday, 25 January 2014 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More house stuff
|
Topic: Stones Road house | Link here |
Call from Stewart Summersby, a friend of Craig Mayor's, this morning. He's building a house in Tantaus Road, and he's interesting for a number of reasons: firstly, he is an electrician and an earthmover. Secondly, he's building an energy-efficient house with heating fired by wood, something he has in abundance on the property. Thirdly he has installed a complete off-grid electricity system, mainly run on solar power, but with planned wind and generator backup.
Over to take a look at the place. Unlike us, he's planning the old Australian-style single heater for the whole house. But he uses it for hot water as well, so the only real issue is a separate coil (water circuit) for the heaters. That would make a total of three in the hot water tank: one from the oven, one from the solar panels on the roof, and one output for the heaters. And that seems to be the most expensive part.
Electricity was also interesting. He uses the solar panels (and later wind) to charge 24 770 Ah lead-acid batteries (2 V), and two 1500 W inverters to provide power to the house, which he claims will power the house for 5 days. That's an average of 300 W, which wouldn't even get us started. The total system cost $36,000, less than he would have had to pay Powercor to have electricity connected to his house. He could have had it cheaper, but for him only the best (German quality, none of this Chinese crap) will do.
That's all fine for him. We're in a different situation. Over to the property to take a look. Yes, we have power cables to the pit outside the house, and digging a trench to the shed would be pretty straightforward. Also took a look at the driveway, which promises to be cheap and easy, probably only $500 for the preliminary earthworks and then a little more than that for the aggregate to put on top—only 40mm stuff until JG King have finished building, then we'll finish it.
Stewart thinks that the best thing for the bracken at the west of the property is to scrape the surface and put it in some of the holes in the ground, then get some clean fill for the rest. Probably another $1000 or so.
|
Echidna sighting
|
Topic: animals, opinion | Link here |
While at the back of the property we found an echidna, which for once didn't run away, even when I prodded it with a stick. I wonder if it was brooding an egg, though it's a bit late in the year for that.
|
Groggy not that unusual after all
|
Topic: general, opinion | Link here |
Somehow I'm not a typical kind of person anywhere, something that doesn't really bother me. But there's something about our new plot of land that suggests that I'm just a run-of-the-mill neighbour. We've already met Garry Marriott, the grey-bearded vendor of our block of land, who also lives next door. Today, while I was talking to Stewart, Yvonne met the neighbour on the other side—Graeme, also grey-bearded. That makes 3 grey-bearded Gs in a row. I feel so ordinary.
Graeme told Yvonne some interesting information about the area. As I had suspected, apparently quite a bit of gold was found in the area in the gold rush days. It seems that the whole of Stones Road was lined with pubs, and on our property was a church. That sounds quite plausible, but it's surprising how little evidence there is of it now.
|
Ports pain, yet again
|
Topic: technology, opinion | Link here |
On with building ports today. My ports-try target completed, and I went back to look for the stragglers. Some I don't know about at all. What's mozplugger? According to /usr/ports/www/mozplugger/pkg-descr it's “a small general purpose Mozilla plugin that displays various types of media formats found on the Internet in your browser”. And it doesn't fetch, and the web site doesn't respond. Who cares? Maybe I do, but it's not exactly on the critical path.
Then there's kdenlive, another package that I didn't recognize, though it seems to be a video editor. It died with a broken dependency:
OK, another one to put at the end. Next was pidgin. It died with the same dependency. OK, time to look for XML::Parser. But how? Where's the port? It took me quite some time to discover that it is textproc/p5-XML-Parser. I wish there were a proper cross-reference for these things.
But textproc/p5-XML-Parser didn't want to install: it claimed to be already installed. More investigation. Something had installed two different versions of perl on the system, versions 5.14.4 and 5.16. textproc/p5-XML-Parser was installed for 5.16, and the packages looked for 5.14.4:
What a mess! Tried removing version 5.14.4:
=== root@stable-amd64 (/dev/pts/1) /usr/ports/lang/perl5.14 29 -> make deinstall
Tried again. Same errors. perl 5.14.4 was still there, at least in part! In the end tried forcibly removing all perls and installing the latest and greatest (but hopefully not buggiest), 5.18.2. Then restarted the builds, which didn't die by the time I went to bed.
|
New photo backup disk
|
Topic: technology, opinion | Link here |
One of the things I got with my latest shipment of components was a new disk for my photos; the old one with a 2 TB file system is filling up, so this one has 4 TB. Hopefully disk capacity growth will outstrip my ability to produce images.
Set it up as before, but things didn't quite work:
=== root@stable-amd64 (/dev/pts/0) /home/grog 19 -> gpart destroy -F ada1
=== root@stable-amd64 (/dev/pts/0) /home/grog 21 -> gpart create -s GPT ada1
=== root@stable-amd64 (/dev/pts/0) /home/grog 25 -> gpart add -t freebsd-ufs ada1
=== root@stable-amd64 (/dev/pts/0) /home/grog 26 -> gpart show ada1
=== root@stable-amd64 (/dev/pts/0) /home/grog 27 -> newfs -L Photos -U -a 64 -e 2048 -g 2097152 -h 64 -i 2097152 -m 1 ada1
See below. The line above is incorrect.
That all worked. And then I tried to mount the file system. EINVAL, and a lot of complaints from the kernel:
What's that? It took me quite some time to realize that I had run newfs against the wrong device: /dev/ada1 instead of /dev/ada1p1. I suppose we should fix that so that newfs will only run against a partition of the correct type, probably with the option of a “dammit” override.
| Sunday, 26 January 2014 | Dereel | Images for 26 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports pain continues
|
Topic: technology, opinion | Link here |
After resolving yesterday's perl problem, things carried on building. GIMP started building last night at 19:20, and was still going this morning. The processor is an AMD Athlon 64 3500+, with 525 CPU marks roughly 5% the speed of my new machine. But part of the slowness is probably due to my decision to copy my photo disk across the Ethernet, and the test machine only has a 100 Mb/s interface. That also affected photo processing, presumably because of contention for the disk.
Finally it finished:
After over 20 hours! Clearly that's a result of the change of perl, but it didn't help. Still, it should be sufficient to rebuild webkit-gtk2. That went on for another 5½ hours, and then, after the longest invocation line I have ever seen (6642 characters):
That looks like it could be an NFS issue; I've seen something like that before. But by that time the day was over, with nothing achieved.
Looking back, it seems I've been complaining increasingly loudly about the state of the Ports Collection for well over 10 years now. On 30 July 2002 I noted that we currently had a large number of bogons in the Ports Collection, and a couple of weeks later I noted that “the ports collection is more difficult to use now than it was when the last edition was published”. This was “The Complete FreeBSD”, and the previous edition was submitted for printing on 17 May 1999, when I wasn't keeping a diary. There are many other such references.
Yes, the Ports Collection has dramatically increased in size in the last 15 years, but somehow other free software projects don't seem to have the continual pain that FreeBSD does. Part of this is a mindset issue, I think—people expect to have trouble and have ideas what to do when the problems occur. Maybe with PKGng people have come to this recognition, and the pain I'm having with it are just teething problems. Let's hope so.
|
Bean weights
|
Topic: food and drink, opinion | Link here |
Yvonne wanted to cook minestrone this evening, and we found a recipe that wanted 14 oz of canned cannellini beans. Canned beans indeed! We use fresh (dried) beans! But how many? 14 oz is about 400 g. How much water do beans take up when being soaked? I guessed 100%, so I used 200 g of dried beans. I was nearly right: we ended up with 420 g.
|
Time for more HDR investigation
|
Topic: photography, technology, opinion | Link here |
Yesterday it was windy, and the Bureau of Meteorology promised only light winds today, so I put off my house photos until today. They lied: it was quite windy. High time to find a better HDR solution. I'll start keeping thoughts on this page, though it'll be a while before it's finished. I intend to take this photo as an example:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
It shows significant ghosting on the yellow (Yellow King Umbert) cannas. There are various ways to address that, including masking that area from a suitably exposed component image, but it'll be interesting to see what the commercial products do.
| Monday, 27 January 2014 | Dereel | Images for 27 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports pain: end in sight?
|
Topic: technology, opinion | Link here |
Into the office this morning to see how my ports build was getting on. Not well:
WebCore had built before. Why not now? Presumably because I had dug the perl installation out from under it and replaced it with something else. That's not the correct way to do it, of course, but given the time it has taken, it seemed a reasonable attempt.
So: what to do? Start all over again? Jürgen Lock suggested that I should be able to get PKGng to work with FreeBSD 9.2, so tried that. First I needed to run pkg2ng. And there I found references to perl dependencies that I had blown away along with perl. Only one thing to do. Start all over again, removing all of the directories /usr/local, /var/db/pkg, /var/db/ports and /compat/linux. Next question: which config file? /usr/local/etc/pkg/conf.? /usr/local/etc/pkg/pkg.conf? /usr/local/etc/pkg/repos/pkg.conf? /usr/local/etc/pkg/repos/FreeBSD.conf? /etc/pkg/repos/FreeBSD.conf? In the end we settled on /usr/local/etc/pkg/repos/pkg.conf?. pkg has this nice trick where there's a dummy installation that reinstalls itself the first time it's run. It installed the latest version, and an old, invalid config file! And of course it complained about what it had just installed:
After removing that, things still didn't work:
Why not? An “Error” occurred.
OK, I had installed 9.2 in the hope that it would be less pain. It wasn't, and Jürgen had told me that 10.0 was now working, so back to try that. For the most part he was right. But then I got strange things like:
=== root@stable-amd64 (/dev/pts/3) ~ 9 -> pkg install digikam
=== root@stable-amd64 (/dev/pts/3) ~ 10 -> digikam
And digikam is the correct name (and not something silly like digikam-kde4-amd64). pkg-plist contains:
Tried rebuilding and installing it. Same problem. Finally checked to see if it had hidden it somewhere non-obvious. Bingo!
=== root@stable-amd64 (/dev/pts/0) ~ 16 -> find /usr/local -name digikam
Why on earth did they do that? On my current machine eureka it's /usr/local/bin/digikam. And the idea of putting binaries for specific ports in their own tree has proven so bad in the past that even the X project no longer does it. But it seems that KDE has to do things its own way. I've already commented about their confusing use of common terms:
What's Home? Root? They're inventions of k3b and bear no relationship to the file systems on the machine.
I was wrong there: it was KDE, not k3b. But it seems that there are people who think that this obfuscation is good. Daniel O'Connor commented:
if you are trying to make a system “for the common man” you need to pick things that give the correct associations for those people
But that's not the case. They don't have these associations yet, and abusing existing terminology in this manner doesn't make life any easier for them—quite the contrary when they discover what file system hierarchies are.
But finally things are working. I now only have 4 ports that didn't install, because there is no package for them. None of them are important, so I'm nearly done. Hopefully this pain is now over.
|
E-M1 fast sequential mode
|
Topic: photography, opinion | Link here |
Yvonne wanted me to take some photos of her and Chris Bahlo on horseback today. Time to catch some images of them in Paso Llano, now that I have a camera that can take 10 still frames a second. Off to take a couple, but things didn't work for some reason. I could have sworn that the shutter went off more than once, but then quickly hit the “Busy” display, and finally only stored a single image—in both raw and JPEG formats! Somehow this camera is altogether too eager to record JPEGs.
In the afternoon, tried again. This time things worked, for 27 frames, all of which looked like this:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
But then it appeared to fill some buffer—after all, that's a total of 400 MB. It took nearly a minute to write to the SD card, which leaves me wondering whether the advertised speed (“SANDISK ULTRA SD HC SDHC CLASS 10 UHS-I 16GB 16G 16 G GB 30MB FLASH MEMORY CARD”) is correct.
| Tuesday, 28 January 2014 | Dereel → Ballarat → Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Applying for a planning permit
|
Topic: Stones Road house, opinion | Link here |
Finally our soil tests are done! Into town to talk to Tom Tyler and finally prepare the application for a planning permit.
Well, that was the intention. Arrived at the display home in Sturt St to discover the air conditioner in the garage had failed, and Tom was half dissolved in sweat. Had a look at the documents, both of us for the first time, and found that we needed a site plan with the location of the house and shed. And Tom couldn't do that, at least not until tomorrow. Left with what I had. This is all taking far longer than anticipated, even after taking into account the expectation that it would take far longer than anticipated.
|
More cvr2 pain
|
Topic: technology | Link here |
Once again cvr2 Just Shut Down. Is this a thermal issue? It was very hot again today, even over 30° inside, but the operating system (Linux) didn't say. But then, it seems that it only ever marks time. Possibly it's just the defaults that are so laconic, but the comparison with FreeBSD is particularly noticeable.
In any case, once again it wouldn't come up. Took the opportunity to blow the dust out of the CPU cooler, and tried powering on again. Still nothing. Disconnect disk. Still nothing. Replace power supply. Still nothing. Remove tuners. Powers on. Replace tuners. Powers on.
So what's going on? Is the old power supply defective? On the face of it, it looks like badly seated (PCI) tuner cards. Left the “new” power supply (in fact the one I removed from eureka 16 months ago) in there, but that seems to be a bad idea: since then it has become noisy, so I suppose there's another new power supply due soon.
|
Tanya growing
|
Topic: animals | Link here |
We've had Tanya for over a month now. In the last month she has almost doubled in weight, from 7.2 kg on 29 December 2013 to 13.3 kg today.
| Wednesday, 29 January 2014 | Dereel | Images for 29 January 2014 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Planning permit woes continued
|
Topic: Stones Road house, opinion | Link here |
Spent the morning collecting all the documentation required to apply for the planning permit. That involved getting a quote for a shed, amongst other things. Finally had everything together except for the site plan that Tom Tyler was preparing, and Yvonne brought that with her when she returned from shopping. Called up the council and went through the checklist. One more item required: a Bushfire Management Statement. Where do I get that? From a Bushfire Management Consultant, of course. It seems that I had received a list of consultants from Steph Durant back in early December, but it was in a proprietary Microsoft format and I hadn't been able to read it. Never mind, said Robyn on the phone, she'd send me another one in PDF format. She then sent the same Microsoft document.
Put it through dischord, my new Microsoft box, and discovered that for the whole of Victoria there are only 9 consultants. The closest one in Julie Lee, in Savage Hill Road, Dereel, with whom I spoke last June about subdividing our property. Called up and got an answering machine.
She called back in the evening. My block of land is apparently very complicated because it's only 75 metres wide, and I need at least 150 metres of defendable space on my own property on either side of the house. Whoever said that the BAL was only 12.5 obviously didn't know what he was talking about. I pointed out that it was in keeping with the (apparently official) calculator, and that I had never heard of a distance of 150 metres. But that's why she has a post-graduate diploma in the matter. My property has a BAL of at least 29, and it would take her 4 weeks to prepare. I should beware of people who try to cut corners; she has never had a statement knocked back.
Yes, of course you don't get knocked back if you overstate the danger. But looking at the calculator, I could have forest (the most dangerous) up to 48 m from the house and still have a rating of 12.5. At 35 metres it's 19. Only at distances below 35 metres does it go to 29. In fact the forest is over 100 metres away, and the shrubs 30 metres away are also too far away to raise the BAL above 12.5.
Gradually I'm becoming suspicious of Julie. The first thing I heard from her was a claim last year that Dereel doesn't want a mobile phone tower. Then a week later she claimed that the property next door could be subdivided. That was interesting, because that would imply that our property could be subdivided too. My discussion with her went nowhere; certainly there was nothing to back up her claim. And today her claims that I need 150 metres of distance from the forest on my own property is not backed up by any documentation I have seen, and it would exclude just about all building sites in the state. So: look for another consultant.
|
Fastway couriers: inappropriately named
|
Topic: general, technology, opinion | Link here |
The remaining components for my new computer were sent from Brisbane by Fastway a week ago. They should have been here yesterday at the latest. Where are they?
|
|
||||||||
“Your parcel is currently in transit between our 'Geelong' and 'Geelong' depots.” What does that mean? And why was it misdirected?
Called up the Geelong franchise and spoke to Bob, who explained that the package had been sent to Ballarat by accident, and since Ballarat is not manned, the tracking information shows “Your parcel is currently in transit between our 'Geelong' and 'Geelong' depots.”
And the next item? “Misdirected—we regret to inform you that it appears your parcel was misdirected to the wrong depot.” I didn't get a good answer to this beyond the information that it had been sent, incorrectly, to Ballarat.
I asked Bob to ensure that the package arrives today. He said he couldn't do that, so I then spoke to Steve, the franchisee, who repeated the refusal. Instead it would go out to Ballarat this evening, after 2 days in the Geelong depot, and be distributed from there. He confirmed that the package would then be distributed by another company, who in turn would involve yet another company, and be distributed by them. He couldn't confirm how often they serve Dereel, nor what they were called, and didn't contradict my fear that it could be another week before I get the parcel.
He then suggested that I could pick it up in the Ballarat depot, which would be a possibility, and explained how to get there. It would still not be available until tomorrow. I agreed to have it picked up there.
I then contacted Chris Bahlo, who will be in Ballarat tomorrow. She asked for the phone number. Steve hadn't told me the number, so I called back and spoke to Emma, who told me that the depot had no phone number and was unmanned. She offered to get the delivery driver to call me when he went to Ballarat, and that I could pick up the package from Paperfreight, the company in front of the depot.
I don't need to point out the contradictions here: Bob says it was a mistake to send the package to Ballarat, while Steve said that, after two days maturing in their depot, it would be sent back there. Steve told me I could pick it up at the Ballarat depot, while Emma tells me that it is unmanned. One person also suggested that the package might be sent via Melbourne.
Am I ever going to see this parcel? I'm not even convinced that it is where their tracking system claims it to be. Sent a stiff letter to them, which, despite promises of immediate confirmation of receipt, received no answer. I hate couriers at the best of times, but this one seems to be the worst.
|
Exetel performance issues rectified?
|
Topic: technology, opinion | Link here |
Yet Another support session with Exetel this afternoon. For testing purposes they gave me a /30 direct Ethernet connection, which made absolutely no difference to the performance. And once again they wanted to run tests with axel, which once again produced the same results. About the only difference was that I got up to 1 MB/s with ftp. As I pointed out, the real application that requires good bandwidth is streaming video, and that generally doesn't use multiple TCP connections.
Some discussion on packet tracing: it seems they don't know what that means! They were talking in terms of traceroute and mtr, and I had to explain to them what tcpdump and wireshark were, and how they worked. But clearly they're not in a position to analyse dumps from each end.
They did come up with one thing, though: rate limiting (“shaping”). Previously the rate had been unlimited, which meant that effectively NBN did the rate limiting. And it seems that their “policer” is a little heavy-handed. I had noticed something to that effect last week, where some very fast arriving packets were followed by a significant delay. So he rate-limited it to 23 Mb/s. My speedtest results were immediate and obvious: from a download rate of 11.74 Mb/s, they increased to 22.97 Mb/s, so close to the 23 setting that it seemed obvious. But then he raised the limit to 24 Mb/s, and things changed: 27.74 Mb/s. And then 33.59 Mb/s. So it seems that the “policer” is very primitive, and the speed can be overridden.
Left it at that for the day. They'll put me back on another place in the network—apparently I was too close to the core—but it looks as if I'll still have a direct “Ethernet” connection, though it's not clear that that would make much difference. That'll take them a day or two.
This has taken quite a long time, but I've got to admit that Exetel's attention to the problem shows much more commitment than their reputation would suggest. If we can get this mess sorted out, I'll probably stay with them.
| Thursday, 30 January 2014 | Dereel → Ballarat → Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Bushfire management statements, continued
|
Topic: Stones Road house | Link here |
After my rather negative experience with Julie Lee yesterday, decided to discuss the matter with Tom Tyler of JG King, who confirmed my opinion. There's another bloke in Ballarat, Todd McClelland, whom Julie had bashed last night, but whom Tom recommends—if I want a consultant. His opinion, which makes sense to me, is that you need that kind of consultant when there's a dispute about the BAL. Apparently there are forms available from the CFA, and also on the web, so I'll probably do that.
|
Into town again
|
Topic: general, technology | Link here |
Phone call early in the morning, from Sue at Paper Freight to tell me that my package had finally arrived in Ballarat. In principle I had asked Chris Bahlo to pick it up for me, but based on the problems I've had so far, decided to go in and pick it up myself.
My way took me directly past the CFA headquarters, so stopped in and asked for a Bushfire Management Statement form. Blank stare. They spent some time discussing where I could find it, and in the meantime I went to pick up my parcel.
Picking up the parcel was interesting. “Do you have a docket?”. “No, they haven't sent me anything”. “OK, here it is”. She didn't ask me for any kind of identification.
Back to the CFA, where they still hadn't found anything. In the end they asked me for my email address, and they would send me information. That arrived later:
And that was all. And yes, the URL was doubled. This isn't the first time I've had a brush-off message from Brendan, and I'm not amused.
|
New DIY market
|
Topic: general, gardening, opinion | Link here |
Also on the way to pick up my parcel I found a new roundabout in Learmonth Road, specifically to service a new DIY market, Masters. They look something like Bunnings, and since I wanted to buy some water spray fittings, went in to take a look. It's very much like Bunnings; hopefully the presence of the two companies in Ballarat will encourage competition. Got my spray fittings, but not potting mix, which was either not where they said it was, or not there at all. I'll be back to look round the rest of the market.
|
Exetel tuning
|
Topic: technology | Link here |
Exetel has been investigating alternatives to traffic shaping on my National Broadband Network connection. Today Eroshan called and asked me to run Yet Another speed test. It seems they have put in their own “policer”, which does no better than the NBN “policer”:
| Date | Time | RSP | Server | Ping | Downlink | Uplink | Comments | |||||||
| 30 January 2014 | 17:11 | Exetel | Internode Melbourne | 29 | 13.21 | 4.25 | Link | |||||||
| 30 January 2014 | 17:12 | Exetel | Internode Melbourne | 38 | 10.27 | 4.24 | Link | |||||||
| 30 January 2014 | 17:22 | Exetel | Internode Melbourne | 44 | 13.14 | 4.18 | Link | |||||||
| 30 January 2014 | 17:22 | Exetel | Internode Melbourne | 43 | 13.32 | 4.23 | Link | |||||||
Looks like it could be a few days before they get it running properly. At least they're trying something now.
|
Watch reset
|
Topic: general | Link here |
Reset the time on my watch again. I had intended to do this every 2 weeks, but somehow I keep forgetting. It was 7 seconds fast, a gain of 11 seconds in 29 days, fairly much the same rate as it has been.
| Friday, 31 January 2014 | Dereel | Images for 31 January 2014 |
| Top of page | ||
| previous day |
|
New computer, finally
|
Topic: technology, opinion | Link here |
Spent much of the morning putting together my new computer. The CPU is an Intel Core™ i7-4771, my first Intel machine for nearly 14 years. It has 32 GB of aggressive Corsair memory. I'm using an SSD (128 GB) for the first time, and after some consideration decided to partition it with a 35 GB root file system, 16 GB of swap, and the remaining 68 GB (don't you love storage systems manufacturers' arithmetic?) will be scratch space for photo processing: many of the panoramas I do have up to 90 components, each of which can expand to a TIFF image of 70 MB or so. Just transferring the files takes up time, so it seems a good use of the SSD:
Processors don't seem to have become physically larger in the last 15 years, but the fans certainly have:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Took my time assembling things, helped by the documentation, which has reached a new low. Both the motherboard and the processor came with instructions on how to install the processor. Neither of them had any text, and the pictures contradicted each other. In particular, the motherboard instructions depicted putting thermal paste between CPU and fan, while the fan clearly was already coated with some appropriate material:
|
|||||||||||||||||||||||
So of course there was no mention of thermal paste in the Intel instructions.
They also differed on how to ensure the correct orientation of the processor in the socket. And how do you attach the fan? They don't say, but it seems that you just press the four corner posts into the cutouts in the motherboard. That's much less force than the AMD fans. Hopefully it's correct, but it sits firmly, and the processor seems to be doing quite well with it. I wish things were better documented, though: it seems that the socket is much more sensitive to damage than previous ones (or at least AMD sockets).
Connecting the motherboard to the front panel was a problem: the cables supplied are too short to reach to where they need to be connected on the motherboard, conveniently located in the far corner:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Yes, they can't know where the connector on the motherboard might be, but they know how big the case is, and you'd expect the cables to reach to the end. This one is audio, something that I wouldn't normally take from the front panel anyway, but it's a silly detail.
And then there's the power supply. It's a very new Antec high-efficiency power supply. And it comes with 4 SATA power connectors and 3 PATA connectors. One of the PATA connectors is convenient for the chassis fan, which for some reason comes with a PATA connector instead of one for the motherboard, but the others are completely useless. I had planned 3 disks, the SSD and of course a DVD drive. So I'll have to buy adapter cables.
On powering up, things Just Worked. In the past I've almost always had some (usually minor) trouble on first power-on, but not today. And it's so nice to read:
Tried the usual things: building the FreeBSD userland and kernel. Userland took 25 minutes, the same as what Jürgen Lock got on his new machine, which is almost the same, but the kernel took 7½ minutes instead of Jürgen's 4½. I'll investigate that later, but it's worth noting that my old build box, with an Athlon 64, took “only” 253 minutes, and it's roughly 5% of the speed of the Intel.
Later in the day saw strange information from top:
That wasn't what I had expected. What used up all the memory? Clearly it's available, but even after the buildworld I had over 25 GB of memory free. To be observed.
|
Dassaulted
|
Topic: technology, opinion | Link here |
Call on the phone today from Max somebody, who was calling because I had downloaded the free DraftSight CAD program from Dassault Systèmes last week. It's not clear what he wanted: as he confirmed, it's free. Maybe he just wanted to help, but I didn't have time, and asked him to send mail. I received none.
|
Photo processing with the new machine
|
Topic: photography, technology, opinion | Link here |
One of the main reasons I upgraded my machine was to process my photos better. So that was pretty much the first thing I tried. Tried with last week's verandah panorama. The control point detector crashed! That was autopano-sift-C, which I don't use, so set panomatic instead. I don't have much in the way of comparison timing, but it seemed to run much faster.
The first part of stitching is running nona, which is single-threaded. On an 8 CPU machine it barely made a difference, and there were noticeable periods where it was idle, apparently transferring to disk. But the disk was the photo disk /dev/ada1, not the SSD /dev/ada0. From iostat:
This was done on the SSD. I'll have to investigate what's happening there.
But the real resource hogger is enblend. How did that go? Orders of magnitude faster:
=== grog@stable-amd64 (/dev/pts/5) /photowork/Hugin-build-eureka 39 -> time enblend --compression=LZW -w -f15134x7567 -o 00-20.tif -- 00-200000.tif 00-200001.tif 00-200002.tif 00-200003.tif 00-200004.tif 00-200005.tif 00-200006.tif 00-200007.tif 00-200008.tif 00-200009.tif 00-200010.tif 00-200011.tif 00-200012.tif 00-200013.tif 00-200014.tif 00-200015.tif 00-200016.tif 00-200017.tif 00-200018.tif 00-200019.tif 00-200020.tif
What caused that? Tried running the version on eureka, with the same result. This is almost certainly some library incompatibility, but which of the 56 libraries is it? vigra? Who knows? How about creating a static binary? How do you do that? Can't you just run ld against the dynamic binary and all the libraries? I'm not sure I even want to find out. In the meantime, found statifier, which claims to do it—for Linux only. Tried configuring it, but:
=== root@eureka (/dev/pts/11) /usr/ports/devel/statifier/statifier-1.7.3 75 -> gmake
That's not because amd64 isn't supported, but because Linux calls it x86_64. Went looking in the configuration for references to that, but couldn't find them. Where now? I'll have to think about that. Daniel O'Connor pointed me at libmap.conf, which might be a good first step.
|
Disk space wastage: where?
|
Topic: technology, opinion | Link here |
A few days ago I built a file system on my new 4 TB photo drive and copied the old one to it. The results weren't quite what I expected:
That's nearly 75 GB difference in used space. Where did it go? Compared the output of dumpfs -m and found:
It's just a coincidence that both partitions are called /dev/ada1p1, clearly on different systems. But who would have thought that those parameters would waste so much space? Rebuilt the new file system and copied it again—this time over a 1 Gb/s link—and it still didn't add up:
What's causing that? Still more head-scratching needed.
Do you have a comment about something I have written? This is a diary, not a “blog”, and there is deliberately no provision for directly adding comments. It's also not a vehicle for third-party content. But I welcome feedback and try to reply to all messages I receive. See the diary overview for more details. If you do send me a message relating to something I have written, please indicate whether you'd prefer me not to mention your name. Otherwise I'll assume that it's OK to do so.
| Top of page | Previous month | Greg's home page | Today's diary entry | Next month | Greg's photos | Copyright information |