This DVD stuff still annoys me. I suspected that there was something wrong with the DVD
burner, but how could I be sure? Sure, the canonical way to do these things is to swap
hardware, but that can be dangerous with intermittent faults, and theoretically the software
should give me clues about what was going on. The problem is “now you see me, now you
don't”: the DVDs work in some drives and not in others, or I can mount them but not
play them with mplayer. Tried lots of
things, including installing mplayer on kimchi (failed: one of the
dependencies didn't build because apparently one of its dependencies, part of X.org, was too old), and finally changed over the drives
between kimchi and lagoon (Yvonne's machine).
The drive that I put in kimchi, which wouldn't recognize the disk when
in lagoon, was quite happy with it in kimchi; so there's at least a partial
system software issue here.
Finally gave up and burnt a DVD. Got the same console messages as the other day, but It looks as if everything went fine,
confirming my suspicion that the messages were benign (or maybe just plain misleading).
It's beginning to look as if the old DVD drive is dying. But why should it be so difficult
to find out for sure?
VoIP over satellite
Yvonne went into town today, and I needed to call her. It took three attempts, not because
of the flaky VoIP connection, but because of the satellite delay. Yvonne answered, saw the
message “Unknown caller” and heard nothing, so she hung up again—before
the satellite delay could deliver my voice. We'll need more practice on that.
Emacs function reference
One of the Emacs functions I use to write these web pages takes input from the
keyboard and inserts a Wikipedia reference, for example yesterday's reference
to Aloe vera. I enter “Aloe
vera”, and it generates the text <a href="http://en.wikipedia.org/wiki/Aloe_vera">Aloe vera</a>:. Straightforward
enough stuff, but there's the underscore in the URL. I need a function to replace the space
with a _. Nothing difficult for a programming language like Emacs LISP, but
what function do I use? All the documentation points at functions
like replace-regexp, which does the replacement in the buffer, not in a string.
I've been putting off looking up the documentation for years, simply because the pain
involved is more than manual correction, but today I finally got round to it.
My suspicions were confirmed: I didn't find it in the documentation. Finally found the
documentation via a Google search in a page titled Lisp Lesson: Regex Replace with a
Function. It's
replace-regexp-in-string. Once I know that, I can get the documentation, and the
usage is trivial: the pain was all in finding the documentation.
Generally did a reasonable amount of work in the garden today. Somehow the summer was just
too warm to be pleasant, and we have lots of things to catch up on.
Sue Blake has written a document about creating sourdough starters, as opposed to maintaining them. She
suggests starting with ¼ cup (whatever that
may be) of flour and to inoculate it with a scant teaspoon (whatever that may be) of the previous stage. At best, that would be a
ratio of about 10:1 flour to starter, much greater than the 1:1 generally recommended for
starter maintenance. Her reasoning for this is that that way there will be fewer undesirable
organisms:
By taking a very small sample to use for the next inoculation, we hope to leave behind
most or all of the less desirable organisms and waste products, eventually transferring
only the components that we do want to propagate. As a guide, one teaspoon is enough to
inoculate up to a cup of mixed substrate.
True, but there will also be fewer desirable organisms. That doesn't seem to give the yeast
and lactobacillus any advantage.
And possibly, by reducing the acidity of the dough (“waste products”), it gives
other organisms an advantage. I suppose I should try it both ways, but there must be some
background for the relatively universal recommendation to double the quantity at each step.
More discussion of Emacs documentation today, in connection with LISP functions. Emacs has
quite a nice collection of help bound to c-h, in particular
the apropos-command function bound to c-h a, which allows you to
enter regular expressions to match the function names. But all I could find with the
text regexp related to buffer functions. I've often wanted to look up non-command
functions, but the closest I came was describe-function, which requires an exact
function name. Today Peter Jeremy stumbled over the function apropos, which isn't
directly bound (access via m-x apropos), which does exactly what I'm looking
for for all functions, not just commands. And yes, of course it
found replace-regexp-in-string.
That's not exactly what I was looking for, of course. I've been using GNU info for decades,
but somehow I still can't find things in it conveniently. Still, apropos is
useful, especially since you can get at it with c-u c-h a.
The weather has taken a turn for the worse—cold and rainy—so didn't do too much
in the garden. Did get round to putting together the mulcher that Yvonne bought in town yesterday; a typical ALDI package which required a crowbar to unpack: the cardboard packing had got jammed
in the unit, and I couldn't pull it out. When I did get it out, there was a maze of
confusing equipment there, and some screws that have clearly been shortened to fit:
Finally got round to playing some bassoons again today. First the Sautermeister, then the Heckel. Finally I'm getting the feel of the Heckel, and it
no longer seems quite so out of tune. It's sobering to think that this is the kind of
bassoon for which Richard Wagner
wrote his music. Yes, he had been dead 10 years when the instrument was built, but it looks
similar to the ones built towards the end of his life, and in any case, it was thus probably
a better instrument than the ones he knew.
Photo day again today, and again I tried the 2 photo HDR sequences I started last week.
They worked well enough, except that messing around with the dials makes it easier to
accidentally move the camera on the tripod between the two shots. Didn't discover until
later that I had done this with one of the photos of the north panorama:
This looked for all the world like a missing
panomatic, but it was there, of course. That's what you get for error messages
which don't give you any information. I think what it was trying to say was
“panomatic failed”, not “couldn't execute”.
But what could I do? It would have meant going out, waiting for appropriate light, and
taking another 18 photos. Instead I faked it and used the left-hand photo above instead of
the HDR version that I had wanted to produce. Surprisingly, it didn't look bad:
To my surprise, Yvonne had suggested cassoulet for dinner tonight, so spent most of the day
preparing one. Things are worse than usual: we couldn't even find smoked sausage, so she
bought a bacon hock instead. And, to our great surprise, she found
some confit de canard in town,
so we put that in too.
The confit did have one advantage, though: duck fat, along with a surprising amount
of gelatinous liquid. Melted it, put it in a glass in the freezer, and within 30 minutes I
had them nicely separated and solid. Here after separating the fat (left) from the top of
the gelatine:
That photo was taken with the on-camera flash, but on the table between my two studio
flashes, so tried the alternative with the studio flashes. No difference. I've been
puzzling about this for some time, wondering if they were triggering too late, so took
several shots with progressively slower shutter speeds, finally ending up at 1/20 s:
These were all at the same aperture (f/3.9), but about the only difference was the colour of
the light, which had clearly been influenced by the ambient light. But the flashes flashed,
and at that aperture they should have completely overexposed the images. Are they maybe
flashing before the shutter opens?
Chris and David Yeardley over in the evening, and we pretty much finished off the cassoulet.
The results? The confit tasted really boring. I don't know if that's because
it was in the cassoulet, but it didn't taste at all salty. It's beginning to look as if
it's not very well made. We have some more left over, so we'll try it on its own. And
somehow the bacon hock couldn't make up for the lack of sausage. Still, it wasn't too bad,
and next time we'll know to stock up if we ever find good sausage again.
Also talking about the possibility of moving house—David is half owner of the plot of
land we looked at six weeks ago, and
this is the first time we've had a chance to talk to him. The result of the discussion:
yes, we think that it'll work, due in no small measure to a lot of good will on David's
part. Looks like we'll be busy in the next few months.
Daylight savings time ended today, causing more surprises than I expected. The first was
that my weather software seemed to have stopped working. Further investigation showed that
the data collection was working, but the graphs were either empty or non-existent. Another
bloody gnuplot issue! Grumbling, went
into my script from hell and found the debugging output:
SELECT unix_timestamp(timestamp(date, time))-946728000, inside_temp, outside_temp, inside_humidity, outside_humidity, inside_dewpoint, outside_dewpoint, pressure_msl, wind_speed, wind_gust, wind_direction, rain FROM observations WHERE date >='2010-04-04' AND date < '2010-04-04' ORDER BY date, time;
Not surprisingly, that didn't return much data: the end date was less than the start date.
Looked at the script in more detail—how I hate date calculations!—and found:
STARTTIME is the time at midnight on the day in question.
It took a while to realize what was wrong: today was a 25 hour day! So STARTTIME +
86400 was 23:00 on the same day, and date -r returned the same day. Another
kludge: just add 3600 and it will always return the date of the following day, whether the
actual time is 00:00 or 01:00.
I wasn't the only one who got hit. I also have a cron job which downloads the
local weather
forecast twice a day. What I had waiting contained:
Western District Forecast
Issued at 5:40 am EST on Sunday 4 April 2010
for the period until midnight EST Wednesday 7 April 2010.
...
Forecast for the rest of Sunday
##
Ballarat ## Max ##
Warrnambool ## Max ##
Hamilton ## Max ##
Colac ## Max ##
By the time I found out, they had fixed it, but you'd think they had would have seen
this issue before.
I've been thinking about becoming more active with my investments, and the difference in
brokerage between traditional brokers (up to 2%) and online brokers (0.1%) is certainly
interesting. Read an article today which recommended Bell direct, so signed up for that, once I could.
What a pain nearly all commercial web sites are! The rendering is, as usual,
impossibly broken, they insist on passwords taken from a limited character set
(9999STUPID works, but it refuses my typical passwords, which contain
non-alphanumeric characters), and even parsing the input for my registration brought up lots
of spurious errors (required an “employer”, whose name could also not contain
certain characters, even for people who are not employed):
There were many others, like a text display window with 14 lines of 5 characters each, but
potentially they might consider it to be confidential information. And it didn't even
render well with Microsoft “Internet Explorer”, though it was a little better.
They use unnecessary flash animation that maxes out one CPU and makes the browser react
really slowly. I think I'd rather pay a little more brokerage and get a web site that
works. But does one exist?
End of daylight savings time also means resetting clocks, and nowadays there are a number of
cameras to consider. Turned on my Nikon
“Coolpix” L1, but it went out immediately. Further investigation showed
that one of the two Lithium AA cells was dead. I had bought them as part of a promotion a
couple of months ago, and taken about 4 photos without flash. I wonder how typical this is;
certainly you'd think that they'd go to particular quality control for promotions.
Out riding today with the Yeardleys, the first time David has been on a horse in years.
Darah tripped and may have hurt herself, but her main problem was that she wanted to go much
faster than the rest of us, so we decided to keep it short.
The hibiscus flowers are about 20 cm across, and the Crassula flowers about 4 mm. And once
again the photos bring home to me that you see more in a macro photo than you do in real
life—in this case, lots of White
flies on the hibiscus:
Chris' brother Jonas and his wife Trish (Patricia) were both born in Germany, and they're
now thinking of going back to Germany for a few years. Trish comes from who comes from
Göbelnrod, near Grünberg Schellnhausen, where we used to live. They'll be living in Pohlheim, where I used to buy my computer hardware from Everyware. So tonight they came along to say goodbye. It's been nearly 3 years
since we last saw them, and in the meantime they have a couple of daughters. Yvonne found herself in a situation that she hasn't been in for a couple
of decades now:
Spent most of the day considering the new house. First over to the building site to
consider in more detail how we would lay things out. The site is pretty much rectangular,
and there's a clump of trees and a dam in the middle, where we're thinking of putting the
house. Here's a static extract from the Google map of the area. The house would run roughly north-south through the blue
pin, and the second photo was taken from the position of the camera symbol, looking at about
240°:
The house would fit at an angle to the right of the house. Somehow it's going to be fun
positioning it. Some of the trees may have to go, but they're pretty old already.
Also looking at how to connect power. We had estimated that we would need 200 m of
underground cable, but it seems that the connection point is at Browns Road, directly on the
other side of the Yeardley house, and we would have to make a wide bypass round that house,
so we might end up with 400 m of cable instead. And that still has the potential to kill
the project.
In the afternoon off with David to look at some more display houses. A couple of months ago we had been looking at the Taverner from homesnow, but they're quite inflexible with their
layouts, and they came up with lots of extras that we hadn't expected, so the real price of
the house we were looking at had increased from $167,900 to just over $200,000. In any
case, it made more sense to look at other houses again, so off first to Macarthur Park,
where Yvonne had been taken by the display home from National Builders Group, but they were
closed. Looked once again at the offering from Simonds, and were a little happier with them than last time.
Then once again to Alfredton,
but this time looked at builders we already knew and hadn't visited in February: JGKing homes (which I can't help reading as
“Joking Homes”), Dennis Family
Homes and Metricon. They all had
houses in the same price range, they looked better from the point of view of finish, and
they were more flexible with extensions. Layout was another thing: Metricon had a
“Fortitude 37” (36.6 “squares”—for some reason builders
can't get used to m², maybe because a square (really 100 ft²) is rather more than 10 m²),
which translates to a real area of about 268m². It's out of our price range, but looking
doesn't cost anything. The usage of that space is really wasteful; the plans don't show any
details, but I'd guess that the master bathroom and walk-in-wardrobe would take up about 30
m² of that. And despite all that, the main living area is completely open:
The layout of the other houses also left something to be desired, but we were pleasantly
surprised with them and decided not to visit homesnow again. It wasn't until we got back
home that we realised the difference: the houses we had been looking at all had floor areas
of about 180 to 190 m², and the homesnow house had 240 m² (or 304 m², as the brochures say;
strangely, they don't mention squares). Given the better layout of the homesnow house, I
fear we're going to have to think yet again. But first we need to know what connecting
power will cost.
It's been a while since I had a satellite outage—in fact, over two weeks—but we
had another one today. I suppose I can put up with that level of outage, but hopefully it's
not an indication that things are getting worse again.
Up bright and early this morning to call up Powercor about power connections to the site of the new home. It took 15 minutes to
get an answer, which was basically: “Go to the web site http://www.powercor.com.au/, click, click, click, click...”. Like so many
“serious” web sites, they don't have easy to remember URLs (this one proved to
be http://www.powercor.com.au/Electricity_Networks/Powercor_Network/Connections/Extension_Works/). Went there and found a number of PDF documents, including a Rural Residential Extension Works form that, despite what Sandra said, I couldn't
submit online, and it crashed my printer when I tried to print it.
Spent some time working out where the
cable would go, all 470 metres of it. Whichever way you look at it, that's going to
be expensive. Decided to go and talk to Ray Nottle about it in the evening. As I
suspected, he would in fact probably be involved in the matter. What he said wasn't
encouraging: at that distance, they'd need 240 mm² cross section, which would correspond to
about 0.113 m³ of copper, or almost exactly a ton of it. The market price for copper is
currently $US 7960 per ton, so by the time it's in cable, it'll probably cost us over
$20,000 for the cable alone.
Things don't stop there, though: since the cable goes over somebody else's property, we
don't have the option to do any of the work ourselves, or even choose the company who does
it. It must be Powercor, and they would put in a high voltage cable and transformer. We
have an indicative price for that in their indicative power line construction costs PDF: $64,000 for 250 m, including the
transformer. The transformer by itself costs $21,100, so it's easy to calculate a price of
$172 per metre, or a total of nearly $102,000 for the total length.
But that's not even the total! In addition, we'd need legal costs for putting in rights for
Powercor to the easement on the Yeardley's property. Ray thinks in the order of $5,000, but
by now, who's counting?
And what do we get for that? A 40 A supply! That's less than 10 kW! We can barely run the
air conditioner off that, and things like continuous flow water heaters, still popular in
Germany, are completely out of the question: in Germany, they typically use 24 kW by
themselves. Why such a low limitation? But in any case, it no longer makes any sense. Our
original idea of subdividing the property would have effectively given us the land for
$30,000 and the electricity for $5,000. Now we're looking at a combined price approaching
$200,000.
While on the Powercor site, found an interesting page: a list of current outages. I'll have to keep an eye on that.
Trying out the Simonds web site today.
They have an interactive house design application which
uses Adobe flash. The problems
running flash under FreeBSD are just too
painful, so decided to try it on cvr2, my Ubuntu-based CVR instead. That didn't have flash installed, and of course the plugin
finder (the one that doesn't even tell you what it's looking for) couldn't find it either.
Finally found a package and tried to install it. It failed:
=== root@cvr2 (/dev/pts/5) /recordings 19 -> apt-get install flashplugin-nonfree ...
Setting up flashplugin-nonfree (10.0.45.2ubuntu0.8.10.1) ...
Downloading...
--2010-04-07 12:01:16-- http://fpdownload.macromedia.com/get/flashplayer/current/install_flash_player_10_linux.tar.gz
Resolving fpdownload.macromedia.com... failed: Name or service not known.
wget: unable to resolve host address `fpdownload.macromedia.com'
download failed
The Flash plugin is NOT installed.
OK, not an issue (well, not a big one): for some reason I often get DNS timeouts. Try again:
=== root@cvr2 (/dev/pts/5) /recordings 20 -> apt-get install flashplugin-nonfree Reading package lists... Done
Building dependency tree
Reading state information... Done
flashplugin-nonfree is already the newest version.
0 upgraded, 0 newly installed, 0 to remove and 320 not upgraded.
Huh? Tested it, and no, it wasn't installed. “Removed” the package and
reinstalled it, and this time it worked. But what's wrong with the Debian package system
that it claims something has been installed when the installation failed? It looks as if
the first part succeeded, and there was no way to register that the configuration had
failed.
The Simonds application is horrendously slow, and it only shows you part of the
information; the rest is on other pages. Tried to open a second window, which it wouldn't
do by itself, so opened one manually and navigated to the web site. And everything I did in
the one window automatically carried over to the other! That's really an extreme case of
“There can only be one”. In the end, gave up; it's just too painful.
After finishing, I found hundreds of messages on the screen from which I had started the
browser:
ALSA lib confmisc.c:768:(parse_card) cannot find card '0'
ALSA lib conf.c:3513:(_snd_config_evaluate) function snd_func_card_driver returned error: No such file or directory
ALSA lib confmisc.c:392:(snd_func_concat) error evaluating strings
ALSA lib conf.c:3513:(_snd_config_evaluate) function snd_func_concat returned error: No such file or directory
ALSA lib confmisc.c:1251:(snd_func_refer) error evaluating name
ALSA lib conf.c:3513:(_snd_config_evaluate) function snd_func_refer returned error: No such file or directory
ALSA lib conf.c:3985:(snd_config_expand) Evaluate error: No such file or directory
ALSA lib pcm.c:2196:(snd_pcm_open_noupdate) Unknown PCM default
ALSA lib pcm_hw.c:1429:(_snd_pcm_hw_open) Invalid value for card
This appears to be a combination of two things:
The application makes noise of some kind, something that was mercifully spared me.
cvr2 doesn't have any functional sound hardware—it doesn't need it. But
the browser isn't taking “no” for an answer.
Some day I must start a series of pages about sites that I have visited.
Over to the Yeardleys this morning to fax our application form to Powercor. During the discussion, David made the
point that the property doesn't belong to us. If the Yeardleys were to apply for
permission, Powercor would no longer insist on doing the work themselves. So in the end
they did the application (it's free). That should make it cheaper, but I still don't see it
being cheap enough to be viable. Just the cable would be at least $20,000, and the
alternative of a high-voltage cable, which would be cheaper in itself, requires a
transformer costing $21,000. I'm beginning to think that this is just too expensive.
My carbon dioxide cylinder is empty already! I only refilled it 8 months ago. I must still have a leak; time to take a more careful look at the
lines. Getting a refill proves to be a problem too: the people are all out in the field
until Friday.
I'm still trying to find out how to enter non-standard characters. I've configured a
“Compose Key” which works fine for Roman-derived European (“Latin”)
characters, though I'm still not completely clear about which characters I can generate with
it—all the documentation (mainly web pages) that I've found gives me examples. For
example, Compose " a gives me ä, and Compose - d gives me ð. But I have
to guess that, since the files I've found don't seem to be reliable: some key combinations
don't work, though it looks as if they should. And it's limited to Roman: how do I generate
a Greek Σ, for example? Simple: I enter the HTML entity Σ. That
works on this web page, but not elsewhere. And my real issue today
was IPA characters, far beyond anything I've
seen documented.
Went out looking specifically for this sort of thing. I still can't find any good
documentation of X's compose function—there's no man page, and the closest I can come
to it is Xutf8LookupString(3), which in the web version is very sloppily marked up:
The possible values returned are:
...
XLookupNone
No consistent input has been composed so far.
The contents of buffer_return
and keysym_return are not modified,
and the function returns zero.
T}
That T} is the end of some tab markup, and the other lines are broken the
way the are because they're table elements. More importantly, though, the man page told me
nothing. The lack of documentation appears to be complete.
Still, found other leads, one of which recommended something built on SCIM, the Smart Common Input Method platform, which is in
the FreeBSD ports collection. The web page
has no documentation. Installed that, and—NO documentation beyond a
relatively long message with a list of other ports which base on it. Looked in the source
tree: yes, there is documentation there, but Once Again the FreeBSD porter chose not to
install it. Has everybody gone mad? It would seem so:
<grO0gle> GRRR!
<grO0gle> ANOTHER bloody port that doesn't install the docco.
<grO0gle> Is this becoming standard practice?
<MavvieRVBD> yes.
<MavvieRVBD> next!
<grO0gle> Have you gone crazy?
<MavvieRVBD> yes
<MavvieRVBD> next!
Finally found a useful general introduction to the concepts on the SIL
International web site. That makes good reading, but it doesn't bring me much closer
to being able to do my own input. I'm beginning to understand people who use Emacs macros
to do it.
The background for this is the horrible mess that people—almost certainly the
British—have made of romanizing
Aryan languages,
notably Hindi and friends. How do you
pronounce Punjab? According to the Wikipedia page Punjab_(India), it's pronounced
/pənˈdʒɑːb/. The Wikipedia page Punjab_(Pakistan) doesn't specify
an IPA pronunciation, but it links to a sound file that
pronounces it /pɑnˈdʒɑːb/. And that's closer to the Afghan transliteration Panjab.
I've known about this inappropriate choice of vowel for some time, and generally I've
understood the errors, but in this particular case I've always pronounced it /punˈdʒɑːb/.
Other words are more obvious: the Hindi word
for cumminseed is usually
spelt jeera, though it's relatively clear from
the Devanagari spelling जीरा that it's
pronounced closer to /dʒɪrɑ/, and that thus a transliteration of jira would make more
sense.
The tomatoes continue to ripen. The cherry tomatoes are doing really well: they're
plentiful and in good condition. The others aren't doing as well: the Rouge de
Marmande have borne almost no fruit, and like the Roma tomatoes, they're attracting lots
of pests, including white fly (which I can attack) and millipedes, which I can't:
The tomatoes aren't the only things that have been attacked by worms. I'm baking tomorrow,
and while preparing the starter, found a thread of stuff in the rye flour, about 10 cm long:
I can't identify anything in it, but I assume that these are freshly hatched eggs of
Indianmeal moths. In general,
though, I think I'm winning the fight against them; every day I open the infested containers
and remove any larvae before they have time to pupate, and the numbers are getting fewer.
I clearly didn't pay enough attention to the graphs I “fixed” when daylight
savings time finished: my graphs were off by 2 hours. Why? Took a look at the scripts and
found:
# 2000-1-1 0:0:0 UTC, the gnuplot epoch. Or so it should be, but for
# some reason my plots come out offset by 2 hours. Use GNUFUDGE until
# I find out why.
GNUFUDGE=7200
Y2K=946684800
These values are in seconds; Y2K was the time_t timestamp at the beginning
of gnuplot time, and all values are in
seconds. There they are, the two hours. So it had something to do with DST. But why 2
hours? Set GNUFUDGE to 0 and—I was off by 4 hours!
Set GNUFUDGE to 14400 and things work again, sort of. But where is this stuff
coming from? It clearly is related to DST, but why is it off in increments of 2
hours? It's high time to rewrite this stuff in a form that's more easily debugged.
I still don't understand (electronic) flash exposure! Today I saw an insect crawling
through a pelargonium flower, and
took a photo of it with my ring flash attachment. The first image was too dark, so I tried
again with manual exposure. Result: much darker:
Yes, the aperture was marginally less, and the shutter speed was at the flash speed. I
assume that most of the exposure was with ambient light. But the flash is supposed to
compensate for that, but I needed to adjust by +2EV to get a reasonable exposure:
That's not the only issue I have. Yesterday I took the photos of the mess left behind
by Indianmeal moth larvae. They
were marginally underexposed, but nothing serious. Then I pointed the Ashampoo optimizer at the photos, and it changed them
completely (“optimized” on the right):
Which is better? The “optimized” version shows much more contrast, but it's
a pretty inaccurate rendition of the original. I really don't know which to use.
Did a little gardening work in the afternoon, not nearly as much as I should have done.
Went out past the verandah just in time to find
an Echidna disappearing underneath it.
Out to get my camera, but they're pretty fast, and by the time I got back it had disappeared
under a mess of tomatoes. I don't suppose they're some of the things eating the tomatoes.
We clearly have something that's been attacking supplies in the garden shed:
The box on the left is snail pellets (non-poisonous to animals), and that had happened some
time ago. The one on the right is seaweed emulsion, in a quite stable plastic bottle. What
could it be? Discussed on IRC, and currently the impression is that it could have been
some kind of Possum, though I'm surprised
that it could have made such a mess without knocking anything over.
Also did some baking. We now have a little butane torch for flambéeing food, so tried an
approximation to
Hannoversches Gersterbrot, where you
torch the surface of the bread. For comparison purposes, but also because it didn't look
quite right, I only did half:
Unfortunately I left the oven on grill. I normally start like that, but then I change to
fan only. If I leave grill on, the surface gets too dark, as it did this time, completely
obliterating the difference of the flaming.
This was also another experiment with trying to get the dough to rise more. I did the
starter in 3 steps instead of two, and used 600 g in the third starter instead of 500 g.
Result: as far as I can tell, no difference. This seems to be as much as the starter can
handle.
Still more modifications to my procedure for HDR weekly house photos. Two weeks ago I dropped from 5 automatic photos to two manual ones, with
the disadvantages that it took longer between the shots and there was thus serious blur, and
also the heightened possibility that I might move the camera on the tripod between the two
shots. But those experiments also showed that I didn't need more than 2 EV between the two
shots, so today I tried again with 3 shots bracketed 1 EV apart, which the camera can do by
itself. I set the exposure to about 2/3 EV overexposure, so the -1 EV was in fact -1/3 EV,
and the +1 EV was +1 2/3 EV. Then I discarded the first shot (“0 EV”). The
result was a drop between the time of the shots from several seconds to about 0.2 s, with a
very visible reduction in blur.
Still, took 2.5 GB of photos even so, and for the first time ever I've taken to discarding
properly exposed photos. After all, I never wanted to take them in the first place, but it
was the only way to get what I want. I think I need to stop putting the partial results on
the web as well; there's just too much stuff.
Our Gruß an Aachen rose has now recovered from the wallaby attack and is blooming again.
Took a photo of a single bloom. Well, that was the intention; in fact I took 8, in the
process gradually coming to a better understanding of the issues. I've already established
that any serious flash photo of this kind needs to be taken with manual exposure, so started
with f/11, which overexposed. Next, f/22 underexposed; but it also gave too much depth of
field, so turned the flashes down to their minimum value and took another photo at f/5.6:
That was even brighter than full power at f/11! It's as if the flashes were still firing at
full power. And indeed, it seems that they were. Normal “automatic” flash
units always charge the capacitor fully and use relatively sophisticated electronics to stop
the unit firing when enough light has been emitted. This has the side effect of having more
power available when it's needed. By contrast it seems that the adjustment on my studio
flash control how much power goes into the capacitor, and the tubes always discharge the
capacitor completely. So when you turn it down, the first flash is still at the levels of
the old setting. Tried again at f/11, f/8 and f/5.6 again, and the last one gave me the
exposure I wanted:
The last image has exactly the same exposure as the first image in the previous row.
Somewhat heartened by this, went to check the validity of my hypothesis that the flashes go
off too early in slave mode. Here the exposure at f/22 again, first with a flash cord, and
secondly triggered by the in-camera flash:
Clearly the studio flashes played no part in illuminating the second photo; the shadows show
that, and it looks like the distance was a little too far for the on-camera flash to
illuminate it completely at that aperture.
In summary, a number of discoveries:
Flash photos: enlightenment
The automatic on-camera modes of Olympus cameras are tailored to delivering fill-in
flash, and as little of that as possible. The result, even for macro photos, are very
wide apertures, frequently full aperture. To do anything useful I need aperture
priority modes, and then I might just as well use manual.
I now have some guidelines on the relative power of maximum and minimum settings. The
“good” photo was taken at minimum power and f/5.6. The photo taken at f/11
and full power was overexposed, and the one taken at f/22 was underexposed. Adjusting
the image under ufraw showed
1 EV difference, so the image should have been exposed at f/16. This means that the
minimum illumination of the units is 3 EV less than the maximum, or 12.5%.
The typical aperture for my macro table and full power flash seems to be f/16, or maybe
f/11 or f/8 for extreme close-ups.
Apertures are measured in “f stops”, the ratio of the iris diaphragm
opening to the focal length f. But the exposure isn't really about that
relationship, though no book I've read in the last 10 years mentions this. It's
really the relationship between the diaphragm opening and the distance of the optical
centre of the lens to the focal plane, v (I learnt the underlying
equation as 1/f = 1/u + 1/v, where u is the distance from the
subject to the lens. Wikipedia refers to u and v as S1
and S2). For most photos, f
and v are close to being the same, but for macros they're not. At a 1:1
magnification ratio, v = 2f, so f/8 is the same as v/16; thus the need
to open up at close distances.
The in-camera flash (or the Mecablitz, for that matter) triggers the studio flashes too
early and is thus useless in conjunction with them. I must find a way to turn off the
slave function on them.
I've been keeping an eye on the hops growing between the east garden and the garage, but
today discovered that the hops growing on the west wall of the garage were going brown quite
quickly. Time to harvest the rest. Got as far as the Pride of Ringwood, and harvested them
all. One of the bines had been so heavy that it snapped the wire it was tied up on, and
carried on bearing hops close to the ground:
Another brief power failure last night. I
was recording three films on cvr2 at the time, and all three failed.
Why? cvr2 didn't go down. It looks as if the real problem was very bad weather
degrading the signals to the point that they couldn't be received, and maybe the cause of
the power failure that was related to the weather.
The other day I received a renewal “special offer” from Money Magazine: renew my subscription for only
$45.00, a “41% saving!”—over the individual copy price. What's the normal
subscription price? They don't say, so I went looking in the current issue. They didn't
say there either, but there's a special for $44.95. They don't say “new subscriptions
only”, so I went off looking at the web site, which is run by a separate
company, Magshop. It took me about 2
hours.
First problem: I already have a login on the site (http://money.ninemsn.com.au/—what have I let myself in for?)—but my
subscription isn't linked to it. I set up this account years ago with regard to a different
magazine, and I had insurmountable problems with it then,
too. They seem to have improved their site since then—the rendering wasn't as
broken—but it's still a pain. There still seems to be no way to link a subscription
with an online account, but now I had the option of typing in my subscription number and
creating a new account number. Tried that, using the same email address, and was
told to log in with the existing details. Catch 22.
While searching the site, came across their terms and conditions. Some
of them are horrifying:
... you ...
11. Consent to (a) the receipt of emails from Magshop containing information and offers
about products and services available on Magshop in which you may have an interest, and
(b) the receipt of emails from publishers of magazines available on Magshop, containing
information and offers about their products and services
12. Acknowledge that these Standard Subscription T&Cs may be modified from time to
time, and agree that you are responsible for regularly reviewing these Standard
Subscription T&Cs, and you agree to any such modification.
That's plain stupid, of course. Nobody can make me agree in advance to any changes they may
make. And the agreement to receiving spam is Just Plain Wrong: they have a check box to
accept it. It's just another indication of the stupidity of so many web sites.
Finally ended up creating a new account with a new email address and linking like that, and
finally things worked. I was offered the opportunity to renew my subscription with the same
special offer that had been in the letter:
So the special offer they gave me cost $0.05 more than the standard subscription, and
the other “special offer” was the same price as the normal subscription. And
this is a subscription to a magazine to show you how to save money! I had already noted
that a two year subscription cost exactly twice as much as a single year subscription, but
that changed with the other specials:
You really need to wonder what they're using for brains. They also allowed me to put two
different subscriptions at different prices into my “shopping cart”:
They mention two different prices: $44.95 in the text, and $45.00 at the bottom of the page.
They also addressed me in second and third person in the text. I'm amazed.
More hop picking (well, plucking) today. Processed about half the hops I cut down
yesterday, and ended up with another 550 g. What am I going to do with them all? Also came
across some of the strangest looking hops I've ever seen. They're the same giant size as
the ones I looked at yesterday, but they have leaves growing out of them:
It's been windy again lately, and this morning I woke up to find a couple of structures
blown over: the protection round the Japanese cherry tree, which bent the trunk, and the
“Eiffel Tower”, the structure on which we have
planted pandoreas and morning glories:
One of the morning glories was partially uprooted, as the last photo shows. Put it back in
again; hopefully it'll survive, though it probably only has about 2 months to go.
Fortunately the pandoreas weren't affected.
More hop plucking. I now have finished the Pride of Ringwood, well over 1.5 kg before
drying. Now for the Hallertauers and the rest of the Tettnangs. Decided to shred the bines
in our new ALDI mulcher, which proved unequal to the task. This is what came out the
bottom:
Message from David DeTinn suggesting that I can get a subscription to “Money”
magazine cheaper elsewhere, for example for $10 per year from Amazon. Indeed I can, but it's not the same magazine: there's also a US magazine
with that not exactly original name.
Strangely, Wikipedia didn't have any
reference to it.
Last month I got a phone call from RACV, our
local automobile club, with whom we have insured our cars. I was told that the insurance
renewal for my car was overdue, something that is very unusual—I make a point of
paying on time, but this time I couldn't recall having received an invoice. Further
investigation showed that Yvonne's insurance was overdue too
(they're both on the same day). So far, so good, apart from the missing invoices.
But then the consultant said I could pay by credit card if I would give her the details.
I'm reminded of the kind of scam mail that we all get every day “Your PayPal account
has been suspended. Please fill out this form and tell us the password”. What's the
difference? This was on the phone—that makes no difference. And this time the
consultant was genuine—that also makes no difference. It's this kind of
approach which encourages people to give their personal details to anybody who makes
unsubstantiated claims of identity, as long as they sound plausible. I expressed my
protest, including that I would change my insurer if this happened again, and asked for a
written explanation.
Last week I actually got a written reply. It's not clear whether the writer had received
the complaint correctly and had not understood it, or whether it had been transmitted
incorrectly, but the answer was completely off topic: he seemed to think that I was unhappy
with being contacted at all, and didn't mention the security impact beyond enclosing a
privacy charter which didn't address this issue. I've been putting off replying, but today
I started. And didn't finish.
The problem is: how do you make people understand? On the one hand you hear almost daily
cases of scams on the Internet, of bugs
in Microsoft operating systems, of crackers breaking in to even big operations like Google.
But where's the “safe transacting”? My bank has publicly accessible security alerts—but it's inaccurate, or possibly they have seen the light and
changed their practices. The bank safely online page states:
Neither ANZ nor the police would ask you to disclose PIN’s [sic] or password
information.
But a little over a year ago this is
exactly what happened to me. Maybe they think that it's safer over the phone, but that
would really be not thinking at all. Again: how do you make people understand? I get the
distinct impression that the people I talk to about it think I'm some kind of crank.
Arguably they're right, but that doesn't alter the fact that something needs to be done
about this. And surely somebody else has already written a document that is more compelling
than what I can do on an individual basis.
From 18 March 2010 to 6 April 2010 I had an
uninterrupted network connection, and I was gradually coming to the conclusion that the
problem had been solved. But then things changed, and since then I have had another 6
dropouts, three of them today. I've contacted the support people, but haven't had a reply.
More harvesting today: finished plucking the hops, and put another 1.15 kg of Pride of
Ringwood in the freezer. There must be a total of 1.9 kg in there already, and I have about
another 500 g still drying. What am I going to do with it all?
Also harvested more tomatoes—the Romas are coming in more numbers now—and
potatoes, which the millipedes have also infested:
Did some reading, and came up with two things: first (this is from a German book), hedgehogs
eat millipedes (but, it seems, echidnas
don't), and secondly, this kind of damage to the potato is due to slugs, not millipedes; the
millipedes can't get in to the potato by themselves. And yes, there are slugs there, so
next year I'll have to be a little more aggressive in my treatment.
The coming hacker barbecue looks like being the barbecue to end all barbecues; at present it
looks like we'll have 21 people, compared to the previous maximum of 10. In addition, we
have more dietary restrictions, and most of the people are coming from New South Wales, an
average of about 1000 km, so many will stay overnight. Spent a lot of time on IRC
discussing things. Sue Blake has an Akai
EWI4000sEWI, which appears to be an
electronic saxophone. Spent some time trying to work out how to connect it to an amplifier;
it seems that the normal connector is a ¼" phone plug, which I thought had become
obsolete (along with the unit of length) decades ago, and which I've never seen as an input
connector before. Discovered that it will also take more modern connectors, so this cable
should do the trick:
In the process discussed the exact dimensions of this kind of plugs. There are three
different sizes, sold in Australia as 2.5 mm, 3.5 mm and 6.5 mm. But the last are sold in
Germany as 6.3 mm, and in fact I measured them as 2.45 mm, 3.44 mm and 6.27 mm. But why do
they still use the big ones?
The impending barbecue at the weekend is gradually getting the better of my lethargy, and
spent some time tidying up the place and otherwise preparing for the arrival of people,
which starts tomorrow. The place certainly won't be spick and span, but people are less
likely to trip over things, and we should have enough beer.
The weekend of the Hackers Barbecue started early with the arrival of Sue Blake in Ballarat
at about 9:30. Had the sudden realization that we would have 10 people for breakfast, so in
to pick her up and get some “ordinary” breakfast food. What do people eat for
breakfast? It's one of those national secrets, and it's been so long since I've had
Australian-style breakfast that I didn't know what to buy. Sue was no help: she's a
vegetarian and currently not very well, so she can't even eat the things she likes. Ended
up buying some plain bread for fussy kids and a few more eggs for Peter Jeremy's cakes, then
off to look for ghee, which has become very
expensive. Some discussion about whether there's a difference in taste between traditional
ghee and the clarified butter that you can buy in the supermarkets, and which they now label
ghee.
Also to Bunnings, who were supposed to
have some pellets to combat millipedes, but didn't find them. Instead found a spray which
claims to handle
millipedes—Carbaryl, which I
thought had been prohibited years ago. Still, desperate problems require desperate
measures, so I bought some.
On the way home, found some interesting roadside plants, most of which posed an
insurmountable challenge to the AF system of my Nikon
“Coolpix” L1. This one appears to
be Epacris
impressa, Victoria's
state floral emblem:
Back home, started cooking some food for tonight. Not only is Sue a vegetarian, but the
other two guests for the night, Jashank and Peter Jeremy, are Friday vegetarians, so we
decided on Indian food, despite reservations: Peter's wife Usha is Indian, and they eat a
lot of Indian food, probably better than what I could serve. Ended up making Alu Masala, Kali
dal and spiced green beans, and also
started making some Masala Vada and Alu Tikka for tomorrow.
That was enough work in itself, more than I remember (this always happens to me). In the
middle, Peter and Jashank showed up, having been through
the Great Ocean Road in record
time, so decided to postpone tomorrow's stuff to tomorrow morning. Fortunately the weather
has picked up—the highest temperature we have had all week was 17.6°, but today we had
26.3°, and it was quite pleasant on the verandah:
The food seems to have done the trick too: on our way back from Ballarat Sue had said
something to the effect that she may not eat anything all weekend, but in fact she really
got stuck into the Kali Dal. The others also seemed to like it, though I wasn't completely
satisfied myself, but I suspect that we could have done with more rice.
And so to the day of the barbecue, the biggest we have ever held:
Things started off with fun. Jashank had pointed out to me that the display on the internal
unit of my WH-1081—something I never look
at—was indicating “battery low” on the external unit. Out before
breakfast to change the batteries, which are conveniently jammed in so that you need a
screwdriver to get them out. Things worked fine, but the “low battery”
indication continued. Then it occurred to me: I had put in some brand new rechargeables,
which are nominally uncharged. Back out again, put in some others—and I got no
response from the unit. 20 minutes of debugging: was it the battery? The outside unit?
The cabling? Things were helped by the fact that I had a second inside unit that I could
take outside with me. Put in a second outside unit with the same result, and finally
discovered that you (normally) have to reset the inside unit when you change the batteries
in the outside unit. I wonder if the thing generates a random code, and the inside unit
latches on to the first one it finds. That would be a more honourable explanation of the
lack of a channel setting.
That proved the validity of my approach to data storage, though: I had a whole lot of
outside temperatures of -0.1°, and I was able to set them to NULL with a couple of
SQL statements. Now if I could find a way to convince Wunderground to change the stored data.
As expected, of course, the numbers gradually dwindled: Sam Lawrance chickened out at the
last minute, and Edwin Groothuis came alone, and did not intend to spend the night—our
10 people for breakfast tomorrow have dwindled to 6, for whom we bought 2 loaves of bread.
We had planned three tables for the main barbecue, including a quarter of the old conference
table from the Kreditanstalt für
Wiederaufbau,
I had left some of the cooking until this morning, and Peter also had a couple of cakes to
bake. Ended up doing my stuff on the “dining room” table:
Callum Gibson and family (the “Mongolian hordes”, his wife Jenny and kids Emma,
Owen, Alice and Rebecca) showed up in time to help with the final preparations:
Callum had also obligingly stepped forward as Grillmeister, though we weren't able to
satisfy his request for tasteless aprons with fake breasts on them. I thought bushfire
protection overalls more appropriate, anyway:
Our worries about feeding and seating everybody were unfounded: Callum's children preferred
to stay indoors (little Owen was afraid of the bees outside, but I saw him carrying a very
patient Piccola around inside), so we ended up with “only” 15 people outside,
and even had a couple of seats empty. Others who showed up were Juha Kupiainen, Lawrance
Stewart (a bloke from Melbourne who is currently working on TCP congestion control) with
Warren, a friend of his whose surname I forget, and Peter Ross, a German whom I knew from
some mailing lists, but I didn't know that he has been in Australia for 8 years, and is
currently in Melbourne. He came by bike from Ballarat.
Peter didn't seem to recall my activity on the German lists. He came into the house
(kitchen) with a few bottles of German beer in an attempt to spread some of the more
pleasant aspects of German culture, and found us (Yvonne,
Chris Yeardley and myself) speaking German. I wish I had taken a photo of his face.
People have been talking about my flaky satellite connection for years, and IPStar didn't disappoint them:
Start time End time Duration (seconds) from to
1271467723 1271468277 554 # 17 April 2010 11:28:43 17 April 2010 11:37:57
1271468470 1271468661 191 # 17 April 2010 11:41:10 17 April 2010 11:44:21
Date Outages Duration Availability Date
(seconds)
1271426400 2 745 99.14% # 17 April 2010
Any improvement I had in the middle of last month seems to be over: the average availability
since I've had the service is 99.12%. And my ISP has sold its satellite operations, so I
fear I'm in for more bad service.
The beer went down well, of course, and Callum finished a keg. We went and looked at the
old hardware—originally the reason for the barbecue was so that Alastair Boyanich
could pick up the PDP-11 (under the antenna in the first photo), but Callum thought that he
should take the Control Data Cyber 910 as well (second photo):
Interesting box—it's a network accelerator, as near as I can put it. Edwin
demonstrated the operation, and sure enough, it worked as expected, accelerating the
download of an MP3 from 30 kB/s to about 3 MB/s—by caching. That in itself is nothing
world-shattering—Squid can do that. The difference here is that the appliance (can't they find a
better term?) recognizes common data wherever it turns up, and then saves it to disk. That
sounds very much like what Alan Kennington and I were talking about years ago:
At the time I had told him that it was crazy to put a disk in a network stream.
The real issue is how difficult it is to demonstrate something like that. You need to show
speedups in situations where current software can't cope. And from my point of view, I'm
interested to hear the implementation details, whether it works like rsync or like blocklets.
People were interested in the beer as well, of course. I'm not brewing anything at the
moment, but I showed the ingredients I had. And Jenny looked up to the ceiling and found it
infested with Indianmeal moth pupae. Off with a can of fly spray to hopefully kill
them. They're pretty well wrapped, and I should remove and crush them:
Most people left round dusk, and in the end we were the same group as last night with the
addition of Juha, who had been offered a bed at Chris' place, but he decided he'd rather
sleep on the floor here than find his way over there late at night. But then he remembered
his cat allergy and ended up sleeping in his car.
Before that, however, we spent some time playing around with cameras on the verandah:
Tried to connect up my Mecablitz 40 CT to Juha's camera. Problem: he doesn't know the
maximum flash trigger voltage for his camera, so I decided to do it with my radio-controlled
remote trigger. That doesn't work properly either: it has enough effect on the flash that
the charge light goes out, but apparently not enough to fire it. I wonder what's wrong
there.
With all the fun yesterday, didn't take most of my garden photos, so did that today, with a
couple of people in tow with the intention of taking photos of me doing it:
Getting my panoramas sorted out with so many people around was a bit of an issue, but it
occurred to me that we should try a specific panorama with lots of people in it:
It looks like Catmint, but it doesn't
have the scent, and the flowers have long since faded. Peter tells me that the orange spots
are Harlequin beetles, though that term seems to be applied to many and varied beetles.
None of the ones I've found look like these.
Spent the rest of the afternoon processing photos—with the help of Jashank, Juha and
Peter I have the largest collection I've ever had for a weekend, a total of 335 images (not
including the components of HDR images), and nearly 6 GB of data:
=== grog@dereel (/dev/ttypa) ~/Photos 753 -> for i in 2010041[678]; do ls -1 $i/*.jpeg | wc -l; done 13
159
163
=== grog@dereel (/dev/ttypa) ~/Photos 754 -> du -smc 2010041[678] 525 20100416
2039 20100417
3377 20100418
5941 total
Processing the “Mongolian Hordes” panorama was fun, and not completely
successful:
The cannas growing out of Juha's
head reminded Yvonne of
a Botticelli painting, though she's
not sure which. But I wonder how to manually frob the individual images to get rid of that
kind of artefact.
Uploading photos wasn't made any easier by extreme network congestion and more satellite
outages:
Start time End time Duration (seconds) from to
1271547788 1271547876 88 # 18 April 2010 09:43:08 18 April 2010 09:44:36
1271550444 1271550481 37 # 18 April 2010 10:27:24 18 April 2010 10:28:01
1271590286 1271590390 104 # 18 April 2010 21:31:26 18 April 2010 21:33:10
Date Outages Duration Availability Date
(seconds)
1271512800 3 229 99.73% # 18 April 2010
I didn't get everything uploaded until nearly 18:00.
We still have lots of leftovers, of course. Called the Yeardleys over for dinner and
finished the marinated meat and some leftover cold sausages. We still have a number of beef
and vegetable skewers and far too many sausages (about 25). But things could be worse.
Still more satellite problems today. I think we can safely say that only part of the
problem has been solved. Clearly we have a serious issue here, and I'm seriously concerned
about what's going to happen when SkyMesh take over. I don't have any reason to believe that they're worse than average, but I don't
know if they'll go to as much trouble. Called up support and spoke to Ashley, who promised to
look in to it. I still think it's time to explore the alternatives again.
Apart from that, Peace! The weekend was fun, but it was tiring, and it's good to be alone
again. Spent some time tidying up and assessing the leftovers from the week, including
things that other people have left behind. What do you do with self-raising flour? The
deep freeze was already full to the gunwhales, so took up Chris' offer to store my hops.
That made a worthwhile amount of space in the deep freeze:
In the evening, more Indian vegetarian food. Peter Jeremy had left behind
a dosai mix, so tried that, with only
limited success. They wanted the mix to be put into the pan cold, which made it not cover
the oil. I ended up putting the whole mixture for two dosai in at once, which didn't taste
too bad, but it took forever to make. I wonder if you can make them without oil in a
non-stick frying pan.
The weather over the past few days has been ideal—sunny, not too much wind, and
temperatures in the low to mid 20s. We were able to eat both breakfast and lunch on the
verandah, and afterwards I finally got round to mowing the lawn.
Didn't quite get finished: it started to rain, not much, but I was looking for an excuse,
and that was it. In to watch TV, in the middle of which it started raining so loudly that I
had to give up. And stopping the mplayer failed: I couldn't access dereel.
In to the office. dereel was still running, but the UPS was screaming. Out to the
switchboard to discover that the
RCD had tripped and
couldn't be reset. Quickly grabbed an extension cable from the garage and connected the UPS
to the circuit in the lounge, then out through pouring rain to look for the problem.
I didn't find it. Ended up disconnecting everything from the circuit, but the RCD still
tripped. At the same time we had water coming in through the ceiling—I wonder if
there's a connection there. Yvonne went to the Yeardleys and
came back with some more extension cables, so now the house is full of them:
The rain itself was pretty spectacular: between 16:20 and 17:50 we had 32 mm of rain, 5% of
the annual total, though my weather software only registered 25.1 mm—I need to look at
that. We had hail and flooding in the garage:
Still no power in the western side of the house. I wonder how long it will be before the
moisture dries out—if, indeed, it's moisture. I can imagine a very high electrician's
bill if I call one out, so if it can solve itself in a couple of days, I'm prepared to put
up with that. But how long should I wait?
Into town today to meet with Peter O'Connell, and did a couple of other things at the same
time. My office chair has served me well for years, but the cushioning is wearing thin, and
the gas lift cylinder is leaking, so time to buy another one. To OfficeWorks, where a surprisingly active
salesperson sold me pretty much exactly what I was looking for. A little shorter in the
back than my present one, but I confirmed that my back never touched above that height
anyway.
To Dahlsens to buy a couple more extension
cables. I don't like the place, but it didn't seem worth going half way across town to the
competition just for that. My fears were justified, though: the price markings were vague,
and when I finally got to the checkout I discovered that my $25 cable suddenly cost $46.
The cashier called for somebody to check the price, not what I wanted: I wanted to know
which was the cheapest. But nobody came, so I gave up. I wonder how that company stays in
business.
The power fault doesn't seem to be going away. Investigated the socket outside on the
verandah, which had been subject to a lot of splashing, but it was completely dry, and I
measured infinite resistance (well, more than 2 MΩ). It wasn't until later that it
occurred to me that I might have been measuring the wrong side of the switch, so out to take
a look at the switchboard. Apart from a dead spider, there wasn't much to be seen:
Measured the resistance there, and got wildly varying values in the range 80 Ω to 400
Ω—corresponding to a load of between 130 and 650 W, far more than I'd be happy
to handle. Over in the evening to talk to Ray Nottle, who gave me the name of an
electrician and also the quote for power for the “New House”: about $40,000 to
$45,000. I think we can put that one to rest.
Back home, found more evidence of what must be going on:
There's a clear trail where water has drained out of the right-hand switch. That's in the
area where the power failed and where the water came in through the roof. I suppose it
would be worth going up there and taking a look, but the ceiling is 3.6 m high, and after my
last episode looking for cables in the ceiling, I have
had to promise Yvonne not to go there myself any more.
Some time ago I noted that the Microsoft Security Response Center
signs its messages with PGP. But recently
there have been bulletins, purportedly from Microsoft, which are not signed. They must be
forgeries. One message has all
the normal boilerplate (word wrap is original):
The Microsoft Security Response Center (MSRC) uses PGP to digitally
sign all security notifications. However, PGP is not required for
reading security notifications, reading security bulletins, or
installing security updates. You can obtain the MSRC public PGP key
at
https://www.microsoft.com/technet/security/bulletin/pgp.mspx.
But there's no signature. The other
message omits even this text.
This is serious stuff. What's Microsoft doing about it? Nothing, it seems. If this is a
forgery, it's a very well done one. The headers look identical to those coming from
Microsoft, and the content matches what's on the Microsoft web site. What's going on? I
can think of two scenarios:
Some sophisticated blackhat is sending out duplicates of the Microsoft security
messages. At some later time, when people are used to them, he will send out incorrect
information with an intent to breach security.
Microsoft management has discovered that their security people are using third-party
security software, and have instructed them to stop doing so.
The fact that there are no signed Microsoft security bulletins any more suggests that
the second scenario is the case. But what stupidity! That's exactly the kind of lack of
understanding that encourages security breaches. Thank God I don't use Microsoft.
More work on finding out the problem with
the RCD today, and in the
course of the day had some success. The first point was that there are two separate
circuits on the RCD, so I disconnected one:
Those loose wires are no danger, since they're not connected to anything. And yes, the
other circuit worked, and included nearly all the power points affected. The only ones
non-functional were in the kitchen and laundry, and on the north side of the “dining
room” which leads to the verandah. Even the other side of the north room, where we
eat, has power.
The location is not overly surprising: that's where the rain came in. So it seems
reasonable to assume that there's a connection. What next? Called up CJ, and he came over
to climb around in the ceiling. I defied the prohibition and poked my head up there too.
Discovered a number of things, some of them surprises:
The manhole only lets us into the part of the roof over the dining room and the south part
of the kitchen. There's no way to get into the roof section over the extreme north of the
house.
There's a valley in the roof line above the beam where the water came through, between
these two roof sections:
The one on the left is the part we could get in. It's reasonable to assume that that's
where the leak comes from. Inspection showed that the roof sheeting (normal profile
galvanized iron) was nailed to a batten, but that there's nothing between the wood and
the ridges. It's pointing down, so that's not normally an issue, but the combination of
high wind and heavy rain can drive water through—exactly the scenario I was
expecting.
Part of the roof area was damp, but there were no junction boxes in the area. There was
a bird's nest, however. It seems that the cables go from one power point up to the
ceiling, across and down to the next power point. That makes sense: that way any
problems can be isolated relatively quickly. So that's probably my next task.
We've planted a lot of less common herbs in the garden over the last couple of years, but we
haven't eaten many of them. Today we had some left-over cold chicken, and decided to use
various herbs for a Vietnamese-style meat
salad. I can't claim that it's authentic: I started with a recipe for something
different and simplified it to a point where it's easy to make and use for using up
leftovers. The results weren't bad; next time I'll take photos.
I've been ranting about low display resolution for some time now, but it seems that other
people are getting fed up with it too. Somebody pointed me at an article by Pete Brown, a Microsoft Developer
Division Community Program Manager specializing
in Silverlight (yes, I had to look up
what that was). He blamed HDTV for the situation, but of course the problem has been there
for years, and it's because of the current implicit assumption that point size is the same
thing as pixel count. Increase the resolution of a display and you reduce the text size for
a given “point size”. Increase the “point size” and you overflow
boxes in HTML pages specified in pixels. The only thing HDTV did was to give manufacturers
an excuse to shrink display size from an already too low 1920x1200 to even lower 1920x1080.
The site allows comments—all I read were in agreement that the display sizes were far
too low—and so I left a comment there. Pete responded noting that he should probably
check his own CSS for that kind of mistake. It's interesting that he hadn't thought of it
before, but of course he's still powerless against the assumptions made by current versions
of HTML. Wouldn't it be nice if people could gradually remove the equation of point sizes
and pixels? The discussion has also spilled over into Slashdot.
I've already noted the problems I have had with Petunias and Lobelias over the past season. Last weekend I forgot to water the last
surviving Lobelias a single time, and they've taken it to heart:
Within a couple of hours, they were completely gone.
But what's that Eucalypt doing there in the first place? This was supposed to be a garden
with small shrubs, and this thing is already 3 m high. Checked back and found that we
planted it on 23 December 2007, so it's only a little over 2 years old.
And I described it as
a Pultenaea daphnoides,
which it most definitely is not. We're still discussing whether to pull it out again; it's
not as clear-cut as with the others on 30 November 2009.
Earlier this month we went out riding, and
Darah tripped. We weren't sure if anything had happened, but today Gordon, the farrier,
came and trimmed her hooves, and found evidence of an injury. Left the left forehoof,
right the right hindhoof, first from above, then from below:
I've already noted that my flaky satellite network connection is bad again. The link
dropouts are possibly fractionally fewer, but now network congestion seems to be getting
worse. Last week it took hours to upload all the photos we took, and this week it took 20
minutes to synchronize my diary; normally it takes about 1 minute:
Sun Apr 25 11:40:19 EST 2010
...
sent 1467254 bytes received 144527 bytes 1454.02 bytes/sec
total size is 698600583 speedup is 433.43
Sun Apr 25 11:59:06 EST 2010
Let's hope that things work better with SkyMesh. As Juha Kupiainen pointed out, the name can easily be corrupted to
“SkyMess”, and indeed it has already been
used, though I wouldn't condemn SkyMesh based on that report. Other threads on
whirlpool look a lot more positive.
Anzac Day today, not for the first time
of course. It's basically a military celebration, and there are things like dawn services
at the Australian War
Memorial, and even at Gallipoli.
But they're making more and more of a fuss of the matter. Why? Is this a grass roots
thing, or are people in high places trying to rouse patriotic spirit to make it easier to
recruit soldiers? It's not just patriotism; nobody makes nearly as much fuss
about Australia Day.
More playing around with the electricity problem today. I started with the hypothesis that
there's a fault in one segment of cable, and that they're linked in a daisy chain, almost
certainly with branches. So the isolation method would be to break these daisy chains and
see which side the fault remains. That lends itself to a kind of binary search, if I
know where the branches are.
The topography is also more complicated than I thought. When I isolated the fault back to
one of the two cables connected to the circuit breaker, I was happy because I thought we had
only about 4 power points connected to it. In the course of time I discovered that they're
a total of 11. And disconnecting a power point is non-trivial: they're in uncomfortable
places, and most of them have been painted in, so disconnecting them can tear the paint. In
the end decided to add another assertion: the problem is probably in the area that CJ
identified as being wet the other day. And that is almost certainly the segment between the
stove and the laundry.
Removed the power point at the stove—conveniently mounted on kitchen surface wood, and
thus no paint to contend with—and confirmed that yes, indeed, it was a daisy
chain. Disconnected one of the strands, and noted that the fault went on the one side of it:
So that's the good news. I'll do some more thinking about the matter, but it looks as if I
can leave things the way they are and reconnect almost everything. But why do I have such a
low resistance? The power point in the laundry was also dry, and it's pretty clear that the
cable goes exactly from the two power points I disconnected. A bit of rain won't do it much
harm. I think the bird's nest is the issue here: maybe the birds pecked through the
insulation. Irritatingly, we can't get into the ceiling of the laundry to take a look.
We'll have to take the roof off, not a good idea at the moment. But I don't think we should
ever use that cable again, at least not until we've looked at it very carefully.
I've been suffering from sub-optimal network performance lately, but it really seems to be
related mainly to network congestion. Today I uploaded a compressed ISO image to the web
site and was amazed by the speed of my 256 kb/s uplink:
386BSD-1.0.bz2
136105159 100% 56.43kB/s 0:39:14 (xfer#1, to-check=47/49)
sent 136123147 bytes received 34 bytes 56494.37 bytes/sec
Still, things could be completely different next week.
Finished my work on the power distribution, which I had delayed because I had to reconnect
my computers (with UPS) to another circuit while I worked on the defective one. Finally we
have the power almost back to normal. We'll have to wait a while to replace the connection
to the laundry; in the meantime we have a power point on the other wall.
Winter's around the door, and we've had to postpone any idea of moving house, so it's time
to finish erecting the greenhouse. Out to try to puzzle out how the thing fits together,
and took 30 minutes to screw down two rails which hold the roof trusses (on an angle at the
very left of the left-hand photo below). Why do all these holes not quite fit? I'm fairly
sure that I am putting the thing in the right place, but some screws have to go through 3
holes that are up to 1 mm offset, and I needed a fair amount of force (not to mention
cursing and swearing) to get the screws in place. Then we tried to put in the roof trusses:
We're missing about 20 cm length in the beams. How is that supposed to work? We also have
some kind of flap that I think is intended to open at the top to ventilate it, but I have no
idea how it is supposed to fit, and these gussets in the middle (the piece with the round
holes in it) came attached to one of the trusses, and there are holes on the other which
match. The only way that could work would be if the trusses started further up. But then
there would be a gap at the bottom, and the flaps wouldn't work. I wish I could find
erection instructions for this thing.
Also planning plants for next spring. We've decided to go
with lobelias
and petunias again, and we need to buy
some seed. Found some seeds for bright red dwarf petunias on eBay, but of course they didn't say what size they are.
Sent a message to the vendor:
How big are the flowers on this variety?
Got a very quick reply (22 minutes):
Thank you for your interest on our store.
This Petunia large head
are 30 CM high.
Admittedly, eBay has carried the concept of email mutilation to its logical conclusion. It
bloated my one line message to 338 lines of mainly irrelevant junk, making the message
itself difficult to find, and this reply was 510 lines long. But even that doesn't explain
the discrepancy. People just don't read mail any more.
One of the things I did find on an eBay product description was that petunias require
frequent dead-heading, and so out to look at the rather sorry remains of this seasons
hanging baskets. Cut a lot of dead stems off and confirmed that the still-living ones were
long and straggly. It's probably too late for these plants, but to be observed next year.
In the process, though, confirmed what I had been suspecting for a while. The little black
spots that came out of the last pruning effort aren't
parasites—they're seeds. They're about half the diameter of poppy seeds, and I must
have collected hundreds of two different kinds—assuming that cross-pollination hasn't
made them all the same. We'll find out; maybe I should buy some seeds anyway.
It wasn't that long ago that I had the impression that nobody cared about higher resolution
displays any more. But more and more evidence is surfacing that people are gradually
getting fed up with the situation. This is from http://xkcd.com/732/:
There's been a discussion about filters on the Oly-e.de forum. It's amazing
how many people say “Buy the most expensive—it's the best, and barely good
enough”. I deliberately bought the cheapest polarizing and graded neutral
density filters I could find, purely to see if they were any worse. If they were, I
wouldn't have lost much money. But I can't see anything wrong with them. Reinhard Wagner
suggested that cheap polarizing filters show marked colour changes when rotated, but I can't
see that with mine either. About the only difference I can see is that they don't have an
anti-reflective coating. That's critical for lenses, but it doesn't seem to make any
difference to plane glasses.
More trouble with the weather station today; for reasons I don't understand, it started
reporting wildly inaccurate values. The outside temperature was frequently reported at
-0.1°, and a couple of rainfall values were over 10,000 mm. Wunderground reports a couple of readings of 2,539.7 mm, and I also found one of
-78.8 mm. Some of these numbers seem to reoccur: -0.1° and -78.8 in various places. I
wonder where this comes from. I had suspected the USB interface, but today the interior
unit showed 4-digit rainfall (that's as high as it goes). It wasn't the same, but quite
possibly there was truncation at some point. The real issue is how to catch these things,
and how to get some semblance of accurate information.
Sent the vendor a message, and he apologized, offering to refund 10% of the price, which I
suppose is reasonable enough. But as he said, the packet was unopened, so it's unlikely
that he was pulling a fast one. He thinks it could be due to drying out, but it seems
unlikely that it could dry out that much. Maybe the original vendor was the one who pulled
a fast one—how many people weigh their spices?
Back to look at the greenhouse again. It looks as if the rails are really intended to be
attached about 25 cm above the bottom of the roof. I can try that tomorrow, but how's that
supposed to work?
John Marshall pointed me at more information about the apparently forged Microsoft security bulletins. As I had suspected, it's
genuine—or at the very least, there are more indications that it is. An entry in a blog claiming
to come from the Microsoft Security Response Center states:
The bad news is that PGP signing is not working correctly in the new system so the mailers
going out today announcing our security bulletin release will not be signed.
That is mind-boggling! This is the security department of the world's largest software
maker, and they can't even work out how to sign their own email! Somebody should offer to
set up a FreeBSD system for them free of
charge. And they further discredit themselves as security experts by publishing the
information not on their web site, but in a blog hosted on a domain not obviously related
with Microsoft! Apart from the obvious fact that customers don't have much chance of
finding the information here, who is to say that it's genuine? The blog is hosted on
technet.com, unlike the security
center, which is logically enough hosted on microsoft.com. A little investigation
with whois shows that the registrant for technet.com is in fact Microsoft,
but who is to say that Microsoft is responsible for the content? Nobody. Indeed, the
postings all contain disclaimers saying:
*This posting is provided "AS IS" with no warranties, and confers no rights*
My best guess is that the people in Microsoft's Security Response Center understand the
issues and do the right thing, but they haven't convinced the rest of the company yet.
It's difficult to believe in their competence now.
Quite a different matter is the recent compromise of the
Apache web infrastructure. It was clearly
well-prepared, and appears to have taken advantage of commercial software that the Apache
Foundation use. The report is interesting for a number of things:
Apache didn't try to cover up what happened. On the contrary, they explain in some
detail how the attack happened, and what they're doing to ensure it doesn't happen
again.
It shows how complicated such an attack can be.
It shows a down side of using TinyURL: the
offending URL was hidden by an innocuous-looking short URL that doesn't arouse any
suspicions. While TinyURL themselves are clearly not to blame, it's just one more
non-obvious thing about Internet security.
The exploit took advantage of bugs in a commercial product, not free software.
Probably the most interesting thing, though, is the enormous difference between this
incident and the Microsoft security team's inability to sign their own email.
Made some curry puffs this evening, requiring short crust pastry, not the sort of thing I'm
good at. Dragged out the cookbook for
my Kenwood Chef mixer, which had a
section on pastries. But there were no pastry recipes there, just pastry-based recipes. I
did find a recipe elsewhere, though, which has both advantages and disadvantages:
The basic recipe is for 250 g of pastry and requires 175 g flour, 85 g of butter or
margarine, and 30 to 45 ml of water. Alternatively, for 8 oz pastry you take 6 oz flour, 3
oz butter/margarine (not quite the same proportions) and 2 to 3 tbsp of water. That makes between 290 and 305 g or
roughly 10 to 10½ oz of pastry. Can't these people count?
On the positive side, the information is tailor-made for the mixer, and things went exactly
as described. I wouldn't have got that information out of a normal cookbook. I had been
meaning to sell the cookbook on eBay, but I
think that's enough reason to keep it.
The multi-person panorama I took last week was less than
100% successful because of the choice cut points, as I noticed the following day. Today spent some time playing around with
Hugin and discovered a Crop
function which allows me to select a rectangular part of each image. That managed to get
rid of some of the weirder artefacts, but also some of the people I wanted to have in the
image. Here the first attempt followed by today's attempt:
I can't see any more mangled faces, but there are many fewer people in there. I need to try
again, but the whole thing takes a lot of time. It's also a nuisance that you can only specify
rectangular sections; that's clearly not the case with the default. I wonder if it's worth
hacking the code.
The PHP scripts I've used to write these pages have “just growed”. In
particular, adding RSS support—something that I don't personally use—has made
things more difficult, as has the feed for ACM
Queue. The biggest issue is that a diary is, by definition, a daily thing, while
comments in blogs are individual. The result was that, while most ACM Queue entries have a
title describing the content, mine had a date. I kludged around that by putting in a title
if, and only if, only “computers” was selected. And that fell flat if I had two
comments on a day. So today I thought out a way to restructure things so that, for RSS at
least, the date of the entry is almost irrelevant (though it is used to set
the pubDate tag), and the heading is included in the individual entry.
That wasn't without its surprises, of course. What happens if the entry doesn't have a
heading, like most of the ones I have at the moment? I kludged that by putting in the date,
but it's not pretty. In future it looks as if I'll need a title for every entry.
How many people read my RSS feed? It's difficult to say, of course, because on the one hand
RSS aggregators access the feeds at regular intervals, increasing the apparent hit count,
and on the other hand each aggregator can have multiple readers. Callum Gibson seems to be
the one person who always reads this way, and he wasn't happy with the change. After some
discussion it seems that the headings were what annoyed him. He uses Newsfox to read RSS feeds, so installed that. It's
one of these silly paned applications (should that be “pained”?), but in fact it
doesn't work that badly.
In the process, also learnt about “categories”, which effectively are the same
as what I have been calling “topics”—maybe I should change the term, but
internally I use function with names like ontopic () to decide whether to
display a specific entry or not, and somehow oncategory doesn't sound the same.
Still, modified the functions to set category headers, and also changed the default headings
to display the categories rather than the date.
After that, Callum was (relatively) satisfied, but ACM Queue wasn't. I had two entries for
yesterday, and only one of them (the first) was taken. Was that because
the pubDate was the same? Maybe. pubDate is synthetic anyway—I had
set the time to 23:59:59—and it was the same for all items. Another hack, starting at
23:59:1 and incrementing for each item. I'll never have 60, so that would give
discrete pubDates for each item. Tried that, and the RSS validator complained: it seems that pubDaterequires two-digit
values for seconds (and presumably minutes and hours). A sprintf just for that?
It occurred to me later that I could have started at 10.
And then? ACM Queue still didn't want to know. Other aggregators ended up with two copies
of everything, but ACM Queue only had the original entry. Maybe it's because the time was
earlier; we'll see tomorrow.
Yesterday's experiments with the Mongolian Hordes were less than satisfactory, and today I
spent much of the day trying to improve. One problem is that Hugin makes what appears to be an arbitrary
decision about which image to select when it has a choice for filling a particular area of
the panorama. It seems that the best thing to do is to start with as little of each image
as possible. I started by cropping exactly the parts I wanted, as in the first image (under
the Crop tab). That left lots of black bars in the preview image:
I could then widen the individual cropping to remove the black bars, in the process knowing
what I was doing and where the image component was coming from.
Hugin has not one, but two image preview modes. One is called “Fast
Preview”, and the other is called “Preview”. Each offers functionality
that the other doesn't: fast preview can display frames showing which image contributes to
which part of the image (and changes the image accordingly as you move the cursor across):
It also allows you to move the image around (“Drag”). That somewhat makes up
for the lack of “centre image here”, which is only available in the normal
preview. More experimentation needed.
Finally came out with an image with which I was satisfied—all I needed to do was
recenter it. And after finishing, I discovered another artefact:
On a number of occasions I've voiced my opinion about the Australian Government's
stupid plan to censor the Internet. The good
news: they've shelved it, at least until after the election. Can we now
replace Stephen Conroy
with Kate Lundy? Of course, some
people see this as a ploy to allow Conroy to claim that he has a mandate for the policy if
Labor gets reelected. Hopefully not.
After yesterday's almost-success with the Mongolian Hordes panorama, tried again and was
finally successful—not once, but twice. Piccola had been in on the act, but always at some distance from the people, so she didn't figure
in the version I had been working on. I had also had to remove at least three images of Sue
Blake, so decided to make an alternative panorama with her and Piccola and as few other
people as possible. The results aren't bad, though I had to trim top and bottom of the
second one because I had removed too much detail there:
From time to time, I've had occasional difficulty building FreeBSD userland from source. I'd get a message:
===> gnu/usr.bin/groff/font (cleandir)
===> gnu/usr.bin/groff/font/devX100 (cleandir)
"Makefile", line 6: Could not find /Makefile.sub
make: fatal errors encountered -- cannot continue
*** Error code 1
Stop in /src/FreeBSD/svn/stable/7/gnu/usr.bin/groff/font.
It always happened in this particular directory, and from time to time I worked around the
problem by removing the directory. But today it happened with a fresh checkout
of FreeBSD 7-STABLE. What's the cause? At first I thought it was an environment
variable clash, so off to take a look. The Makefile
in /usr/src/gnu/usr.bin/groff/font/devX100/ is pretty minimal: it consists of only
three lines, one of which includes ../Makefile.inc. And that in turn includes
the Makefile.inc from one level further up, and that defines DIST_SUBDIR.
So what went wrong? Tried running it make
in /usr/src/gnu/usr.bin/groff/font/devX100/ under ktrace and found:
.include "../Makefile.inc"
.include "${DIST_DIR}/Makefile.sub"
.include "../Makefile.dev"
"
6006 make RET read 191/0xbf
6006 make CALL open(0x82234e10,O_RDONLY,<unused>0x1b6)
6006 make NAMI "../Makefile.inc"
6006 make RET open 5
6006 make CALL fstat(0x5,0xbfbfd0d4)
6006 make STRU struct stat {dev=103, ino=6313215, mode=-rw-r--r-- , nlink=1, uid=1004, gid=1000, rdev=25219922, atime=1272596794, stime=987511002, ctime=1269403621, birthtime=987511002, size=162, blksize=4096, blocks=4, flags=0x0 }
6006 make RET fstat 0
6006 make CALL read(0x5,0x8223f000,0x1000)
6006 make GIO fd 5 read 162 bytes
"# $FreeBSD: src/gnu/usr.bin/groff/font/Makefile.inc,v 2.1 2001/04/17 12:36:42 ru Exp $
DIST_SUBDIR= font/${.CURDIR:T}
TOPREL= ../..
.include "../Makefile.inc"
"
6006 make RET read 162/0xa2
6006 make CALL stat(0x82206a80,0xbfbfd11c)
6006 make NAMI "../../../Makefile.inc"
6006 make STRU struct stat {dev=103, ino=6294273, mode=-rw-r--r-- , nlink=1, uid=1004, gid=1000, rdev=25136370, at
There's no chdir there, so the current directory is
still /usr/src/gnu/usr.bin/groff/font/devX100/, and so the second include
of ../Makefile.inc
from /usr/src/gnu/usr.bin/groff/font/devX100/../Makefile.inc should refer
to ../../Makefile.inc. But instead it looks for ../../../Makefile.inc! How
can that happen? With the help of Peter Jeremy and Callum Gibson, I've established a number
of things:
It doesn't happen to Peter or Callum.
It has happened to me numerous times over a considerable period of time.
It happens on different machines and releases that I have.
It happens all the time with the stable/7 tree.
It happens all the time if /usr/src/gnu/usr.bin/groff/font/devX100/ is the current
directory, but not always when I'm in /usr/src/ and do a make buildworld.
So what is it? Next step is to debug make, I suppose. I'm really puzzled.
The RSS stuff seemed OK yesterday, except for one thing: ACM Queue still only takes one topic per day. Is it maybe
set to take only one per minute? I need to decide whether to change the pubDate
parameter yet again. The coming change of month might be an appropriate time.
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.
