So today was the first day of spring, at least officially. In fact, it was cooler than
yesterday, and much wetter; nevertheless, the Ballarat weather statistics for August show
that it wasn't quite the wettest on record. There were “only” 161.0 mm,
compared
to an average of 75.3 mm, and a record of 167.1 mm.
The rainy first day of spring reminds me of other first days: ten years ago I started this diary, and I've been keeping it ever since. And on that
occasion too we had rain, and a burst water pump. And 52 years ago I recall the first time
I paid any attention to gardening topics: we drove (I think)
from Melbourne
to Broadford, where my mother
grew up, and stopped by a stream with lots of
blooming wattles. Then, too, it was
raining, and my mother said “It always rains on the first of September”. I
asked why, expecting some profound wisdom, but it was just a generalization.
It seems that the Geelong Regional
Libraries have a mobile library that comes by here every fortnight. Took a look at
the web site, and discovered that I could sign up online—if I work around the massive
breakage. First there's a link to the signup page:
The registration page itself was typical of narrow-minded web programmers: they ask for
BLOCK LETTERS, whatever that's supposed to mean in a web context; I assume they mean
capitals. Clearly the web programmers haven't heard of functions that change cases. And
they want a street name, but no number, and a “Suburb”, whatever that means.
Clearly it doesn't mean town, because one of the choices is “Australian Capital
Territory”:
It doesn't ask for a street number, but it did want a post code, something that should be
clear from the location. So I selected Dereel and left the Postcode field
empty. Bad idea. It complained, and erased the entire form:
I was able to go back and fill out the rest of the form. I stuck to proper use of upper and
lower case, and for the fun of it gave a slightly incorrect post code and a PIN pair that
didn't match:
And it accepted everything! Things that are obviously wrong are the lack of a street
number, mismatching PIN and wrong post code. In addition, it accepted lower case
characters, though it appears to want upper case only. It gave me a code (not shown here)
with instructions of how to modify it for use:
Never mind, I had to give it an email address, so I waited for the notification email. It
didn't come. I've seen some broken web sites in my time, but this one really takes the
cake. The surprise is that the company who did it hasn't tried to hide their name.
So I signed up again, same name, same address, same email, but otherwise with correct
information. Tried to sign in as described, and got an authentication failure. That's
enough: it's just not worth the pain.
Another cool, moist day, and once again I didn't do very much. Cleaned out the filters for
the water pumps, something that I do every three months. In the winter it doesn't seem to
be necessary:
In the process checked the pressure of the pressure cell on the house water pump, which was
too low, but it gave me the opportunity to try out two new ALDI toys, a digital tyre pressure gauge and an air pump.
I'm not sure about the air pump; it runs off a 12V cigarette lighter connector, and it
worked well enough for the pressure cell, but I'm not sure it's any better than the other
toy I have, and it could have significant problems pumping up a tyre, for example.
One of the key features of Wikipedia is the
simple markup, which means that just about anybody can contribute. It's certainly in stark
contrast to things like DocBook, and for
a long time I've seen it as a reaction to the excessive complexity of such markup languages.
But simplicity doesn't mean ease of use. I update a lot of Wikipedia pages, but I have
written almost none, and even updating proves to take excessive amounts of time.
Today I looked at the page for the Nikon
F camera, which was both minimal and inaccurate; it claimed that the Nikon F was the
first system camera, whatever that
may mean. Elsewhere there was much more information in the history
of the single-lens reflex camera page and on the German version of the page, so I spent some
time incorporating that. It took me several hours.
I'm no stranger to writing, of course. I wrote my books
in groff, and I write this diary
in HTML. But I use real HTML markup, not a
wiki markup, and it seems easier. One thing's clear, of course: the transition between
source written, by default, with a web browser form and the displayed page is slow
and clunky, and even though I use Emacs to edit it, it's still painful—one line
per paragraph certainly doesn't help. Even when I was finished I discovered that I had
omitted a space between two words. Why do I go to the trouble?
It's been nearly 2 years since we declared the verandah complete. But it wasn't quite, though for some
reason I didn't note it in my diary: we had tried to connect the new gutter to the house
water system, but gradually come to the conclusion that it wasn't within our ability, and I
should get a plumber to do it. Things weren't made any easier by a blockage somewhere in
the system. But where do I find a plumber?
In Ballarat, and he'd charge the best
part of $100 just to come here. So I did what I always do in these circumstances:
procrastinate.
So, for nearly two years, I've had water just running off from the one side of the verandah,
and all the water from that side of the house running out of the blocked gutters. Finally
I've found somebody, Bryan Jackson, who lives just round the corner. He came over this
morning with his mate Roger and did the job.
Finally things were in place, and all that needs to be done now is to re-route the vent pipe
at the right so that we can push the water pipe against the wall. But I can do that myself.
While he was here, Bryan also finally connected a hot water tap for the dish washer. I had
bought it months (years?) ago and left waiting for the plumber. Now the dish washer cycle
time has dropped from 60 minutes to about 32.
With the drain pipe in place and the blockage removed, I can now attend to the beds round
the verandah, which had been continually flushed by water from the roof. Put in a
significant quantity of compost and finished putting in
a Jasmine vine at the base of the mesh in
front of the garden shed, and planted
some Nasturtiums, which have shown
significant aptitude in climbing up mesh.
I wonder how long it will be before they give up. There are plenty of places in the horse
sheds for them to build nests, and some have already done so.
The ten-year-odd drought seems well and truly broken. Last month we narrowly missed the
heaviest August rainfall on record, and by this evening
Ballarat had had 80% of average
September rainfall. Here in Dereel we had
over 30 mm so far, most of which fell last night. The results can be seen in the dam, which
is frequently completely dry. The previous fullest state was on 10 November 2007 after some particularly heavy rains (first
photo). We exceeded that level last week (second photo), but this week there seems to be at
least 30 cm more water in the dam (last photo):
I had to change the viewpoint for the last photo, because the place I normally take it from
was under water. Also took a photo across what used to be a dry gap between the two halves
of the dam, from the mound on the left of the previous pictures:
The rain continued most of the morning, accompanied by high winds, both of which made it
difficult to take HDR photos, so instead played around with various other ideas, in
particular the consideration how many images I need for a panorama. Currently, with a 9 mm
lens mounted horizontally, I have a horizontal field of view of 88°, but I have been spacing
the images every 30° (12 images for a full 360° panorama). Then last week I tried the
panorama from the north-east of the verandah with vertical orientation (field of view
72°), and also got by with 12 images.
Another issue with the panoramas is that this makeshift panorama bracket is getting on my
nerves. Today the whole thing tipped over: the ball head isn't strong enough to hold the
stuff on top unless it's mounted relatively vertically. I've been considering a bracket
from LinkDelight[URL since rotten],
which looks pretty solid, and the pano-MAXX, which seems to do the job for vertically mounted panoramas, but which
doesn't really look stable enough for my work. But the documentation for the pano-MAXX suggests increments of 45° for the vertical panoramas
and 60° for the horizontal ones. Tried the latter with some of the panoramas, and it seems
to work fine:
After taking the photo of the join between the two halves of the dam, went back down to the
paddock and slipped on the wet muddy surface, landing on my back. The good news was that
the camera didn't get damaged, and that I didn't hurt any joints. The bad news was that I
pulled a muscle in my calf, and spent the rest of the day hobbling around. Chris came along
for dinner in the evening. She's an expert in injuries, and gave me some horse liniment to
rub into it, along with the joyful news that it'll hurt more tomorrow. Looks like I'll
spend even more time in the office for a while.
Chris was right. Woke up today with a particularly painful leg and spent most of the day
immobile. I had planned to do some programming or something else for which I normally can't
find the time, but in the end just played around with the layout of my diary.
What's a good web page layout? Lots of people with more insight than me discuss this at
length, of course. But they all have a disadvantage that I don't: they're not me, and I
want the pages to be visually pleasing to me. This means, at the very least, no distracting
side bars, no clever layouts that can easily render badly on displays that the writer didn't
intend. In short: not a “major web site” like Telstra or Livejournal.
But there are a couple of things that could be improved. Easy-to-read text is no more than
80 characters wide, not because that was the width of punch cards, but because of the focus
limitations of the human eye: it becomes uncomfortable to read when the text is wider. And
on a normal screen my diary renders about 180 characters wide. That's presumably the reason
for the side bars in Livejournal.
So I went experimenting to see how I could do that too. Problem: I don't want to restrict
the width of everything, just the text. I can do that by specifying alternate <div> tags for photos and text, which gave me something like this (shown in
sans-serif for something approaching clarity):
I've since moved to this layout, so this example no longer looks any different. I've
rendered it in grey to indicate that it's quoted. The image display has been faked.
Another issue with the panoramas is that this makeshift panorama bracket is getting on my
nerves. Today the whole thing tipped over: the ball head isn't strong enough to hold the
stuff on top unless it's mounted relatively vertically. I've been considering a bracket
from LinkDelight, which looks pretty
solid, and the pano-MAXX, which seems to do the job for vertically mounted panoramas, but which
doesn't really look stable enough for my work. But the documentation for the pano-MAXX suggests increments of 45° for the vertical
panoramas and 60° for the horizontal ones. Tried the latter with some of the panoramas,
and it seems to work fine:
So it looks as if once again I'll be doing some major changes to the way I do my photos.
That allowed me to use the full width of the screen for photos, but doesn't it look ugly?
Tried various alternatives, but so far I haven't found anything I like, so I've reverted it
all.
I've had problems with TV reception in the
past, but they were typically related to specific channels. Today I discovered that
the already bad radio reception for ABC
Classic FM had deteriorated into noise, and that none of the TV channels presented an
adequate signal for the tuners to get a lock. Hobbled around and got a real (analogue) TV
to see if there was any signal at all. There was, but it was unacceptably poor. So: not
the tuner cards. Antenna? Antenna amplifier? Confirmed that, despite the winds we've had
lately, the antenna is still there and is still pointing in roughly the correct direction.
And removing power from the amplifier resulted in complete loss of signal. The toy antenna
that came with the last tuner was also not able to provide any signal at all.
So: what is it? After the last time I'm hesitant to call the broadcasters. But something's
clearly very wrong. I'll experiment more when I can walk again.
Ate
Wiener Schnitzel this evening.
That's not exactly an unknown dish, but the devil's in the detail. Wiener Schnitzel comes
from Wien (Vienna). That's not a foregone
conclusion: Wiener
Würstchen (Viennese sausages), which the Americans call simply Wieners, come
from Frankfurt am Main; the
corresponding sausages in Vienna are called Frankfurter.
It's almost impossible to get a real Wiener Schnitzel in Germany, for example, just
“Schnitzel wiener Art” (“Viennese style Schnitzel”). In Germany,
it's almost invariably pork, but the original is made from veal. But that's not the only
difference. Almost every recipe that I have seen describes the same way to make it: take
the Schnitzel, put through flour, egg and then breadcrumbs, cool for a while and fry in a
pan until golden. There's some difference of opinion about what it should be fried in;
candidates are butter and lard. But one way or another, it's not a difficult dish.
But they don't do it that way in Vienna. Instead of frying in a pan, they deep fry it.
We've been trying it that way for a while, and I've come to the conclusion that it's
better. In a pan, the breadcrumbs absorb too much fat, and the dish tastes soggy. With
deep frying, it comes out crisp, and it also seems to be tenderer.
Why don't I find this in the cookbooks? I haven't found many in my collection of cookbooks,
but they all use a pan. (Bonniers Kokbok, which, if you believe Bonnier's web site, is no longer available,
mentions deep frying, but it still prefers frying in butter in a pan). The German Wikipedia page does state that the schnitzel should be fried in plenty of fat, but it's still in a pan.
We've had much more than our fair share of rain in the last month and a bit, but it's
showing no signs of abating. Today I hobbled to the dam to see what things looked like
there. It's even a little higher than a couple of days ago (left):
The problems I had with Hugina couple of weeks ago made it clear that I need
to upgrade my system. And that requires recompiling literally hundreds of ports (currently
I count 861 of them registered in /var/db/pkg/), and again that takes days. So for a
long time I've been building new systems in the background and installing to an alternate
partition. The file systems on dereel look like this:
/destdir is a second root file system (including /usr), and that's where I
install the new system. When it's ready, and not before, I simply swap the entries
for / and /destdir in /etc/fstab and boot from the alternate partition.
The problem has been that DESTDIR doesn't always work as desired. I've seen this
with pwd_mkdb, for example: it claims cross-device links.
So this time I'm doing it differently, on a different disk on my test system. When it's
(finally) done and tested, I'll copy the entire partition to dereel. Today I got as
far as upgrading to the latest base system, but then I ran into the afternoon, where the
data volumes are restricted, so I'll continue tomorrow.
It's bad enough having some Heisenbug in my TV recording setup, but the problems I mentioned
normally apparently weren't part of them. Checked on the ABC reception page, but there was no mention of
any problem there. Then, in the evening, reception came back. It seems that it wasn't only
ABC that was affected, but clearly it's another indication that disseminating information
via the web is not high on their list of priorities.
Yvonne decided on
fried Camembert cheese for dinner
tonight. We've had it before, fried in a pan. Served with cranberry jelly, it's not at all
bad. But how do you make it? Like deep-fried milk,
it seems to defy the laws of physics.
Yvonne did some googling and came up with a couple of alternatives: deep-fried or heated on
a sandwich toaster. She decided on the latter as being more controllable. She served it
when molten cheese started to come out of the casing. And it was cold.
So: what to do? Cold Camembert is still edible, but we wanted it warm. So the microwave
oven came to the rescue (20 seconds at 550 W):
As planned, started building the ports on my new machine today. It's supposed to
become dereel.lemis.com, but I can't call it that. Called it dewrong before
Jashank Jeremy offered the more appropriate defake. Spent a lot of time loading
files from the Internet, and from the other end of the world, before I realised two things:
Contrary to my own instructions, I hadn't done my symlink
from /src/FreeBSD/ports/distfiles to /usr/ports/distfiles, so I was
loading files I already had.
defake's IP address is 192.109.197.167, which is part of a block once
belonging to the University of
Dortmund. Much software seems to think that I must therefore be in Germany, and
whatever decides where to get my software from also seems to think so. But I can
override that, as Edwin Groothuis told me some time ago. Just add this line to the
(possibly new) file /etc/make.conf:
Continued like that, and by disconnect time at 13:00 I had installed about 150 ports in 2
hours, 19 minutes—all dependencies for X. I'll hopefully finish building X tomorrow.
Things still seem to be particularly cool for the time of year, but at least it didn't rain
today. A couple of new plants are coming
out: Dracunculus vulgaris
(left) and some shoot that I can't categorize, from a place where I didn't know I had
planted anything. Potentially it's related to the Mexican something that Sue gave me last year, but I think that's
further to the right:
Continued in the morning with building ports for the new dereel. Finally, at 10:10,
finished building X, a total of 263 ports, including both PERL and Python. Total time 4 hours, 6 minutes. Not for the first time, I'm reminded of what I
wrote in 1993:
Even the complete X11R6 windowing system takes only about 4 hours on a 66 MHz Intel 486
PC.
Isn't progress marvellous? Also ran into new problems:
Building the nxml port failed without an error message:
===> nxml-mode-emacs23-20041004_6 incorporated into Emacs 23.
*** Error code 1
Messages starting with ===> are normal informational messages (see below).
But in this case it seems to have been a reason for aborting the port, which caused my
entire build to fail.
Installing the Linux base software failed:
===> linux_base-f10-10_2 linuxulator is not (kld)loaded.
*** Error code 1
This looks like a recursive error message: “I can't build because I don't
exist” masquerading as an informational message. But the
“linuxulator” is a kernel module that provides the kernel side of a Linux
environment. It gets built with the kernel. So why doesn't the port load it? I don't
know, but after loading it manually, it works:
[root@defake /usr/ports/emulators/linux_base-f10]# kldload linux [root@defake /usr/ports/emulators/linux_base-f10]# make ===> Vulnerability check disabled, database not found
===> License check disabled, port has not defined LICENSE
===> Extracting for linux_base-f10-10_2
=> MD5 Checksum OK for rpm/i386/fedora/10/basesystem-10.0-1.noarch.rpm.
=> SHA256 Checksum OK for rpm/i386/fedora/10/basesystem-10.0-1.noarch.rpm
There seems to be no canonical Samba port.
The one I have in my scripts was net/samba3, which happily started building and
told me it was deprecated. So I went looking for the correct one, which you'd expect to
be net/samba. But there is no such port,
just net/samba3, net/samba32, net/samba33, net/samba34,
net/samba4-devel and net/samba4wins. Decided on net/samba34. But
wouldn't it be nice to have an easily recognized version?
Feedback from Chris Yeardley and a couple of others, all saying that they liked the
sans-serif limited width display of my diary better. So I'll leave it like that, at least
for the moment. Feedback welcome. At the very least I need to do something about the day
and topic headings.
In this example, the area to the east
of Beaufort is overlaid with
part of the Dereel map. There's no good way to recover from this, since a screen refresh
returns to the original URL, which is generally in no way related to where I have
navigated. Callum Gibson suspects that this is due to breakage in Google's asynchronous
loading, probably exacerbated by the slow satellite connection. The trouble is that lately
it seems to happen almost every time.
My leg is gradually less painful, and so out to try to prepare the potato bed, which was
sorely in need of attention. Didn't finish: it seems that my leg was too painful, so
back with only half the work done.
Mail message this morning from Camilo Jorquera, the Web & Intranet Coordinator of the
Geelong Regional Libraries web site, in reference to my article on the problems I have had. He told me that the problems had been solved, but that doesn't
seem to be the case. About the only improvement is that the login page is no longer placed
in a too-small window with scroll bars in both directions. The rest remains, in particular
the fact that I can't log in using the code I have been given.
Sent a message to Camilo telling him so, and got a reply back which I suppose illustrates
the frustration of his position: as I had noted, the system was badly broken, but it appears
that there's little he can do to fix it. He blames it on “closed source”, but
really it just a product that isn't suitable for the use to which it's being put (and
probably for which it has been sold). The sad thing is that people get money for this kind
of product.
Continued with my ports build today. At least it's going relatively smoothly, but it takes
forever! About the only problem was installing acroread, which
seems to have some sensitivity to Linux versions. But I'll look at that later when the rest
is done.
My leg is a lot better now, but it's certainly not completely healed. Did a little bit more
work in the garden, but maybe I should wait; it's straining things somewhat. One thing
needed attention: the 1000 litre water tank we have near the bore pump was overflowing,
a problem it has had before: it seems that the ball valve gets out of level. We don't use
the tank anyway, so disconnected it.
Yet another morning spent building FreeBSD ports. Spent almost the entire morning on web browsers—firefox took 2 hours, 22 minutes, and
SeaMonkey, recommended by Edwin
Groothuis, took about 2 hours, including installing yet another interpreted language,
this time tcl. The rest of the time was
spent installing things like Flash and other plugins.
Spring is slow in coming, but some things are gradually coming: the
white Camellia is blooming (and,
surprisingly, not going brown), and
the Salix melanostachys is
getting its distinguishing black catkins:
While I was taking photos, saw still more kangaroos in the paddock. Since my leg still
hasn't healed completely, called Yvonne to chase them away.
They disappeared surprisingly quickly, but in the process Yvonne didn't look where she was
going, tripped over a kangaroo guard on the new birches, fell and bruised her ribs. Now
we're both invalids.
Message from Michael (the one without a surname) today referring to my problems with Google Maps a couple of days ago. If a
page renders incompletely, you can't refresh with Ctrl-R or some mouse click; that
will just take you back to some previous page. Instead you have to click on the Link
link at top right and copy the resultant link into the URL field. Now wouldn't it be easier
for all concerned if the software were just to do that for you, or at least give you a
chance to link without the copy/paste operation? And wouldn't it be good to have a clear
way of supplying information back to Google? In fact, it's there—now:
After considerable examination of the pano-MAXX, I think I should buy one. Some time ago I looked at the web site and
discovered that they don't give prices for postage outside Europe. So on Tuesday I sent them a message asking for the postage to Australia. By
today I had received no reply—why do so many companies ignore email?—so I sent a
reminder. The reply, from a certain Torben
Ohme, was astonishing:
Currently we only deliver within the European Union. Thank you for your understanding.
Their advertising says (in English):
All prices incl. 19% german VAT. Non EU-citizens pay net. prices.
Clearly they mean residents, not citizens. But that's a clear contradiction. It looks as
if Mr. Ohme was taking the path of least resistance.
It's been a week now since I slipped and pulled a muscle in my leg, and recovery has been
slow. There are still lots of things to do in the garden, but I don't think I'm in a
position to do any of them. At least Yvonne's rib bruises
are not as serious, and she spent some time mowing the lawn.
This is the first posting of a new category, history, in this case computer history.
When I started in the computer industry in April 1973, one of the things that fascinated me
were UNIVAC's drum storage. They had only
just started using disks at all (IBM clones, if I recall correctly), and the storage
hierarchy ranged from tape, slow drums
(FASTRAND), fast fixed-head drums, to
core memory. The fast drums evolved in the days when the drum was main memory. By
the early 1970s they had evolved to be used as swap space. The pride and joy was the FH-432
fixed-head drum. Its statistics would almost look good today: it ran at 7120 rpm and thus
had an average latency of 4.24 ms. It also had an impressive transfer rate, but that's
another can of worms.
Transfer speed
When I thought of writing this article, I was sure about the transfer speed. Now I'm not so
sure, and I can't find any definite information. A couple of recollections:
The drums were connected to the memory of
the UNIVAC 1108
or UNIVAC 494 by one or two
channels. Both machines used 750 ns core store with a 1.5 µs cycle time and with up to
fourfold interleaving: there were up to four banks of memory, 64 kW each, and the bank
was selected by the bottom bit (for two banks) or two bits (for four banks). This meant
that successive words were transferred to each bank in turn, allowing up to 4 words to
be transferred in 1.5 µs, a rate of 2.67 MW or 12 MB per second (for the 1108, which had
36 bit words).
I'm not sure what happened with three banks, like we had with the 494 at Karstadt; possibly interleaving had to be turned off
Despite this speed, it was possible for two drums to saturate memory with two channels,
causing overruns. This places a lower limit on the speed of about 6 MB/s. Strangely,
the operating system didn't always handle this situation properly. My first task as a
systems programmer at Karstadt was to modify the drum handler (what we'd call a driver
nowadays) of the 494's Omega operating system to retry overruns. It wasn't even
difficult. I suspect this didn't happen on the 1108 (which I had used as a user-space
programmer a few years before), but it's strange it could happen at all.
On the other hand, a number of references I've found on the web show much slower rates.
One brochure for the 494 suggests that it's done character-wise, and gives a transfer
speed of 1.2 megacharacters/s, about 900 kB/s. Another
reference suggests that EXEC 8 (the operating system for the 1108) takes 2.13 ms
to transfer 512 words (36 bit). It's probable that this includes the operating system
overhead, but as it stands it represents a speed of about 240 kW/s (1.08 MB/s).
What does this say about the data organization? Each track had 1024 words (36 bits on the
1108, 30 bits on the 494; despite the lack of space, the 494 literally threw away space).
At 7120 rpm, this means a single-track transfer rate of 121 kW/s (545 kB/s on the 1108, 456
kB/s on the 494). The 494 had 5 characters per word, so 1.2 megacharacters correspond to
240 kW, and to get the transfer rates in the 494 brochure would require two tracks in
parallel. At least that ties up closely with my calculations (it wouldn't be convincing if
it were 1.73 tracks or 2.5 tracks, for example). The 1108 example also matches this speed
quite well, though it's not clear what happens to the operating system overhead.
So why did people tell me that two FH-432s could saturate memory? Why did I need to fix the
drum handler? I/O had priority over the CPUs, so during such transfers they ground to a
halt (no cache). But you'd need 25 tracks in parallel to achieve 6 MW/s, so the number of
tracks would be somewhere between 25 and 50, clearly in the right range for my recollection
of word-wise transfers. Did things change in the course of time? The 494 brochure appears
to have written at the time the machine was introduced, in 1965. The 1108 paper was
presented in late 1971. My first experience with the drums was in May 1973, and the work on
the drum handler was done in June 1976. Could they have changed the configuration of the
drums in that time? That would at least explain why the drum handler didn't handle overruns
elegantly. One thing's for sure: we only had two FH-432s at Karstadt, and the brochure says
that a minimum of three were required.
So: what speed? 1.08 MB/s or 19 MB/s on the 1108? I like to think the latter. And even
today that's not that slow. But that's only part of the FH-432. It stood in a 2
metre high cabinet and was fed by a special 440 Hz power supply. The engineers told me that
this was necessary because of the extreme high rotational speed, but I suspect it had more
to do with the military background of much of UNIVACs hardware. And it only had 256 tracks:
the total storage capacity was 256 kW (1.15 MB or 960 kB). Even at the slower speed, it was
possible to transfer the entire contents in just over a second. At my 19 MB/s, it would
take 7 or 8 rotations, about 60 ms, to transfer the entire contents.
On with building ports today, with yet another behemoth package. X, firefox and SeaMonkey are by no means the only multi-hour builds. Today installed Project X, which requires Java. I was spared this
emetic manual download of the sources: for some reason there hasn't been a new release since
last time I had to do it. But it still took 3 hours, most of the time until I ran out of
off-peak traffic. I now have 439 ports built; on dereel I have 861, so this is a
little over half way in 6 days.
On with the FreeBSD ports installation. Today
it was mainly graphics and IRC stuff. The big one was pidgin, something that I can't even recall using. It spent over an hour
installing all kinds of GNOME extensions,
apparently streamers and things. Why do you need that for an IRC client?
In any case, it's ridiculous to spend over a week installing software. I'll have to think
of a better way. One that has been suggested is using portmaster to keep a base system
running, and updating from that from time to time. Another suggestion is to use a
jail to build
the system, which at least means that I don't have to run another system.
It's been over a week since I slipped and pulled a muscle in my leg, and I gradually have to
face up to the fact that it's not really getting any better. As a result, even my normal
slow rate of work in the garden has ground to a halt. Tomorrow I'll go to a physiotherapist
and get the leg looked at.
Yvonne did some work in the garden, including transplanting
the volunteer Lobelias that have sprung
up in various pots on the verandah. It's strange: we bought them last year, and they
died pretty quickly, but apparently not before
self-seeding all over the place. But in the process, discovered that my
prize Hibiscus
rosa-sinensis is looking very unhappy:
Pruned first, took photos next and then went looking for the likely cause. After some
consideration, I think it's wind burn. It wasn't the cool weather, because only a few weeks
ago it was looking quite happy and even had buds on it. But since then we've had some heavy
winds. Next year I think we should put these plants in the greenhouse for the winter. Took
one of the prunings and put it in some potting mix; it's the right time of year, and maybe
it will strike.
My leg is not getting much better, so in to town today to see Heather Dalman, a
physiotherapist, something that I haven't done for nearly 9
years. She gave me the pleasant news that I can expect it to take between 4 and 6
weeks to heal. And that
with rms coming on Thursday.
She did massage it and make it feel better, though.
The physiotherapists are just across the road from the Botanical Gardens, so in
to see what was going on. Not much, it seems: even the big round succulent bed is being
replanted, and there wasn't much to see. They had some succulents on sale, though, and
bought an unspecified Aeonium with only
the note “big yellow flowers”. But that's what I want. There's a nice example
in the gardens, a far cry from the flowers of our Aeoniums:
With the help of a helpful and well-informed staff member, also positively identified
our Gasteria as a Gasteria carinata
var. verrucosa “variegata”. Here's a photo of a couple of Gasterias they had on sale. The
one in the front of the pot is the verrucosa, and if I recall correctly, the one
behind (brighter green) is the normal carinata.
Coincidentally, our Gasteria has finally finished blooming after nearly 6 weeks. In that time, the flower spike grew to 155 cm long, and the head
(shown at the beginning on the left) grew to 77 cm:
Continued building ports today, and after half an hour of more gstreamer plugins,
decided I wasn't really interested in pidgin, so removed it from the build and continued with the next, GIMP. Then I went into town, and when I got back it was
waiting in one of its innumerable configuration screens. So I didn't get even one port
installed today, just 22 dependencies.
While in town, sent some books off to Jana Oesselmann in Zeven/Badenstedt, Germany. The
books themselves had cost about $60, and they weighed about 1.7 kg. She wanted them sent
airmail, which cost $67—far more than I would have expected.
In the other direction, I'm still trying to find a source for a pano-MAXX. Karl Grabherr in Wien is
prepared to send me one, but first it has to be sent to Austria, for which the company wants
20 €. Within Germany it's only 4 €. Did a bit of investigation and found that a
1 kg parcel from Austria to Australia costs only 19 €. Somehow the relationships make
no sense at all.
My leg felt a whole lot better after Heather had treated it, but it didn't stay that way.
In the course of the afternoon and evening it got steadily worse, and by the time I went to
bed I could barely walk. That's not what I was expecting.
Somehow my leg isn't getting much better. Woke up this morning with it feeling marginally
better than yesterday evening, but no better than yesterday morning. Kept pretty immobile
all day, and in the evening Yvonne suggested I try some pills
she has for “pulled muscles”; she had removed most of the label to write that on
the package, but I was just able to find something saying
“diclofenac 50 mg”.
That's the stuff that's in the gel I've been rubbing on the leg, and so I thought it would
be worth a try. It's not 100% clear that it made the difference, but from then on my leg
was much less painful.
Making the target config-recursive, which runs all the config scripts in
advance.
Making the target checksum-recursive, which downloads all the necessary
tarballs. That would at least enable me to carry on building in the afternoon, greatly
shortening the time it takes.
Building the ports in a jail on dereel. This would speed up the builds
themselves by a factor of 5 to 10: the CPUs are much faster than in defake, and
there are four of them.
Started moving data across in preparation for setting up a jail. The /home directory
on defake, which contains the Ports Collection hierarchy, was nearly full. Removing
the build directories freed up 7 GB of space in a very varied time:
make clean goes through the tree recursively, and takes forever:
...
===> x11-wm/xfce4-session
===> Cleaning for xfce4-session-4.6.2
===> x11-wm/xfce4-wm
===> Cleaning for xfce4-wm-4.6.2
real 235m53.648s
user 184m2.723s
sys 32m29.195s
The alternative is to find the work directories and remove them:
=== root@dereel (/dev/ttypc) /home/jail/ports/home/ports 16 -> time find . -type d -name work | xargs rm -rf
real 8m33.723s
user 0m1.603s
sys 0m23.218s
That can be improved upon, of course: I don't need to go more than three directories
deep to find the work directories, so I could add a -maxdepth 3 to the above
and speed things up further.
The real issue is: why doesn't the clean target in the main Ports Makefile
do this?
One difference that occurs to me this time: SkyMesh doesn't have a free mirror server, so I have to pay to download the tarballs.
With Wideband and Aussie Broadband that
was different, so I didn't have to limit the builds to the morning. Yet another reason to
want to change from this abysmal ISP.
My leg is—thank God—continuing to improve. I don't like taking medication, and
in case of doubt I won't, but it's fairly clear that
the diclofenac has reduced
inflammation and made it much easier for the muscle to heal. But there's a way to go yet,
and once again I kept pretty still all day.
My enquiries on
various forums brought no useful information at all. In fact, I don't think they even
understood the question. So I decided that, since I was going to throw it away anyway, I'd
take it out of the fridge and see what the “worms” did. Nothing. After about a
week, some mould took over and caused a horrible smell, which was missing before. The third
photo shows that the “worms” have not changed at all.
This ports building is taking an eternity. There are two obvious reasons: the processor is
slow, and I only have about 3 hours a day to actually build. I really should pre-fetch the
tarballs and build in a jail, but for the former I need to do a significant amount of
coding, and for the latter I need to understand jails better, so today I just continued as
before. My free low-peak traffic is running out, so I continued into the afternoon, and by
15:49 I installed the first port of the day, digikam. Then a couple of smaller other ones before starting on hugin, the cause of the upgrade. That's
slow enough to build on dereel; on defake it was glacial, and it stopped when
it wanted to install gcc version 4. I wonder why it needs that.
The Aeonium I bought on Monday was
well-developed and showed a couple of buds coming out of the stem. The staff member
suggested that I should cut off the stem above these buds and plant the rest as a separate
plant, so that's what Yvonne did yesterday. I didn't take
any photos of it when I bought it—something that I should really do as a matter of
course—but today the buds had already grown considerably:
You'd think that somewhere they'd say what it really is
(OK, Cucurbita pepo, but that
includes a lot of things). In fact, it seems that people really do call this
variety Vegetable spaghetti.
It has taken quite some time, but today
finally Richard Stallman made it
to Ballarat. Things weren't all plain sailing. To start with, I wanted to add a link to
the official announcement, which was severely broken:
Investigation showed that it only displayed correctly with Microsoft. Even Apple (as shown)
couldn't handle it. The W3 validator showed
that it had a total of 375 errors, and the breakage was due to the use of (undeclared)
Microsoft-specific characters. Just what rms would want. To the credit of the university,
they took down the page as soon as I told them.
At about 15:35 a call from Jack Burton in Adelaide, telling me that rms was on the train to
Ballarat. That would have had to be the 14:53, due to arrive in Ballarat at 16:18. Nobody
had contacted me, so off in a hurry to pick rms and Dora up. Arrived only a little late,
and of course they weren't there; the people in Melbourne had been rather vague in their
terminology, and in fact they were still waiting in Spencer St. station. Down to the SMB
campus of the university and met Glenn Stevens and some of the other people organizing the
event. The lecture theatre holds 120 people, but 160 had registered, so they were planning
to video the talk and broadcast it live to an overflow theatre.
To do that, it seems, they had to have the projector running—there was no provision to
turn it off while the lecture was being transmitted. As one of the staff said, “We
want it to be as easy to use as possible. The lecturers don't want to be involved with
setup issues”. But of course there's no innate conflict between flexibility and ease
of use. In this case the simplicity was an obstacle to ease of use. The real problem (and
the explanation) is what the projector was displaying: a Microsoft “desktop”.
And they had no way of turning it off short of shutting down the system! In the end, that's
what they did. They scrapped the idea of the second theatre and put chairs in the aisles
for the overflow. So much for “ease of use”; once again, it's commercial
software holding people to ransom. I'm continually amazed by the constraints people take
upon themselves so that they can run commercial software.
Back down to the station, and rms and Dora arrived as planned shortly after the scheduled
arrival time of 17:19. Glenn took some of the merchandise, and rms, Dora and I headed off
to Ballarat Bird World. In view
of the time, rms wanted to skip the visit, but we had asked them to stay open specially for
his visit, so he agreed.
On the way he gave another example of bad language from Vline (itself an
obfuscatory term for the state railway transport system). It seems that the train he came
in on wasn't a train at all, also not a coach, but an “intercity” (the previous
one had been a train). And it continued on to Maryborough, in the process changing its
length and its identity. It seems that only the first two coaches were continuing, so they
made an announcement that continuing passengers should move there. Well, that's the
intention. I had always thought that a coach was a part of a train, but for Vline it seems
it's a bus. They have a new term for coach now:
“velocity unit”. rms wanted
to know whether it was Imperial or metric.
At the bird world, up to see Cheeky and Jesse, the red-tailed black cockatoos (Calyptorhynchus banksii subspecies Macrorhynchus), and rms decided that it
had been worth the trouble after all; he had never been that close to black cockatoos
before:
In the process, discovered that my careful preparations for the evening had been compromised
by the sudden dash to the railway station. I had packed a camera bag with a spare
(telephoto) lens, flash unit, and also brought a tripod. I didn't forget them: I forgot the
camera instead, so I had to take all the photos with my Nikon
“Coolpix” L1, which is irritating me more and more.
Back to the university, arriving at 18:30 for the barbecue, which started at 18:00. That
proved not to be such a problem: I had thought that the idea was for people to meet rms
before the talk, but in fact we all went upstairs to a staff room and ate there. I was not
unhappy about that: my leg, which had been feeling relatively acceptable in the morning, was
getting more and more painful. Clearly it's not up to much walking yet, and I'll have to
scrap the ideas of going to
the Great Ocean Road with them
tomorrow.
In the process, rms made acquaintance with Australian Open Sauce:
He said he was tired—not surprising, since this was the third talk in a little over 24
hours—but he was clearly not as tired as last time I saw
him, where he went to sleep while holding his talk. It seems that it's the only time
he has ever done that, and until tonight he didn't know how long he had been asleep (it was
only a few seconds).
The talk turned out to be a hybrid of “Free Software and Your Freedom” and
“Copyright vs Community in the Age of the Computer Networks”, and as a result
ran far longer than planned. Back home and spent some time talking, while rms ran into
trouble with our lack of DHCP connectivity.
In the end, he had to patch into a switch with cable, something that people don't seem to do
much any more:
It's bad enough that things take so long with only 3 hours a day to build software, but
today I seem to have managed to make things worse. hugin takes forever to build, and somehow I
managed to accidentally repeat the process. As a result, another day with just a
single “base” port (a port that I want, as opposed to a port installed as a
dependency).
Up after far too little sleep, but the good news
(for rms) was that a couple of
crimson rosellas had shown up in
front of the verandah. rms off to try to get photos of them, but without luck. But those
are the first rosellas we've seen here in weeks. Maybe they sensed a kindred spirit.
One of the things that rms states in his “visiting hacker” package (my name, not
his) is that he doesn't eat breakfast. We do, and we weren't sure about Dora. In the end,
we did all eat breakfast, and I got at least one sign of approval for my sourdough rye bread:
Unfortunately, my leg was still hurting, so I decided not to go on the tour of
the Great Ocean Road. Sasha
Ivkovic showed up to take them, and almost before they had arrived, they were gone again.
As a result, didn't get much chance to talk. We did discuss old computer software,
including the origins of the
words CAR and CDR. There's no
reasonable doubt that they stand for “Contents of the A part of register number”
(specifically, the last 15 bits) and “Contents of the D part of register number”
(specifically, bits 3 to 17, the last 15 bits of the first half word), but what does
“register number” mean? Unfortunately, rms was unable to help. I'm pretty sure
it means “memory location”, as the assembly macro CDR appears to
confirm:
LXD JLOC,4 # Load index from decrement
CLA 0,4 # Clear and add
PDX 0,4 # Place decrement in index
PXD 0,4 # Place index in decrement
The instruction mnemonics are from John Savard's IBM 704 page, but that
page doesn't say what the parameters are, nor what “index” means. I need to
look at this in more detail, but it seems that the first instruction loads the decrement
part of JLOC, and the others somehow massage it. It's not clear what the 4 means.
It looks like a (processor) register number, but there weren't that many of any kind of
register on the IBM 704.
Also discussed the issues in MACLISP,
which ran on the PDP-10. Although the PDP-10 was a good machine
for LISP, the lack of
tag bits in the machine word made it difficult to identify the kind of object that a pointer
was pointing to. The 704 had a 3 bit tag field that was used for that purpose, but the
PDP-10 (also a 36 bit machine) had 18 bit addresses, so there was no space for any further
information in the word itself. Instead they allocated like kinds of object on specific
pages, and it was the page address that indicated the type.
This was also one of the reasons why rms chose UNIX and C as a base
for GNU rather
than ITS/LISP: he
couldn't see a good way to implement a fast, reliable LISP system on commodity hardware. I
wonder what the world would have been like if he had.
One of the issues we had in getting rms to come here in the first place was because I had
suggested to Josh Stewart of the Ballarat Linux
Users' Group that he invite rms. But rms is much more resolute about not supporting
just “Linux” than he used to be, and said it wouldn't be appropriate to speak to
a group that concentrate just on Linux. He suggested changing the name of the group to
indicate a dedication to all free software. I was for it, of course, and so was Josh, but
so far nothing has happened. Today rms suggested the use of the
word Eureka somewhere. That
sounds like a very good idea.
On with my ports building, including a few big ports such as Apache, PHP and
Wine. To my surprise, they all compiled, and
all the rest of the ports too: I'm finished. It was rather a surprise.
So where do I go from here? I should still install a jail, but not in /home/jail:
that's what /destdir is for. And then I should finally finish configuring the thing
and boot from it. Mañana, at the very earliest.
Bought some new maps for my GPS navigator on eBay. Well, they're different maps. They seem to be marginally older than the ones I have.
But how do I install them? First I clearly should back up the memory of the navigator with
this emetic Microsoft “ActiveSync” software. The instructions tell me:
Plug the GPS into your computer, it should show as an external hard drive.
I think this means “Connect the navigator to your Microsoft-based computer using
the USB cable. Microsoft will show it as an external storage device.”
But that doesn't happen. Instead, I get a window with very little information:
Open the drive and copy the main folder onto your computer as a backup file.
I have great difficulty interpreting this. If you open a disk drive, the warranty is
voided. And this isn't a drive at all. At best it's a storage device with its
capabilities castrated by ActiveSync or whatever it's called. Clearly this is
bad language. Then there's a main
folder. What's main about it?
Top-level? And what's a “backup file”? How do I copy it?
This device is supposed to “show as an external hard drive”. Microsoft
assigns drive letters to file systems, but I couldn't find one here. About the only
thing that I could select on that window is “Explore”. I don't want to
explore, I want to access the file system. But I couldn't find a drive letter, and
the “My Computer” window doesn't show one:
So I looked at the ActiveSync window. It doesn't give me much choice:
“Explore”. I don't want to explore, I want to access the file system.
But I don't seem to be able to do that. The window doesn't offer a copy function, and
there seems to be no way to access the file system with “standard”
Microsoft tools.
So started copying that directory (it's really the only one that interests me) and
somehow managed to stop it accidentally in the middle. It had copied 500 MB of
data somewhere, but apart from creating a directory structure where I asked,
there was nothing to be found. Repeated the operating, allowed it to complete, and only
then did the files appear. It seems that “ActiveSync” writes the data to
disk, but doesn't create the file entries until it's finished. I wonder how to reclaim
the space from the first attempt.
After all that, copying files to the navigator was easy enough. Copied the files for
Germany and France, but the system didn't see them. Why not? I'll have to investigate some
other time.
Mail from Harald Arnesen about the new diary layout. He likes it; that seems to be the
general opinion in what feedback I have received, so I think I'll keep it. He's
particularly happy that the text is narrower, but computer output isn't. I agree. But he
doesn't seem to have noticed (or cared) that the line length of bullet lists is the same, so
they jut out beyond the nominal text margin. I still need to think how to do that right.
It's gradually becoming clear that just about every photo I take of the garden could be some
kind of panorama. I've changed the dam panorama so that it covers the lagoon as well:
It also covers both of the other dam photos, so I suppose it's about time to retire them.
In addition, considered the long-standing shot south-east from the verandah at the north
of the house. It has shown some of the biggest changes of all. Here the first, taken
on 21 October 2007, and potentially the last, taken last weekend:
I can't make up my mind what to do with this one. It, too, is a candidate for a panorama,
but that changes things completely, especially if I look round to the west as well (second
image):
One thing's clear, though: at 300 mm focal length (equivalent to 600 mm full-frame), I can't
get close enough to the bird. This is the whole background to my search for longer
telephoto lenses, of course. Put on my Hanimex 300 mm f/5.5 with a three-fold
teleconverter, making 900 mm (1800 mm effective) and tried again, but it's not very
convincing. I had to stop down to f/11 to get any depth of field, and the teleconverter
made f/33 out of that, so I had to expose for 0.4 seconds, during which time the bird moved:
My leg is getting a little better, and despite the hesitant coming of spring, there's plenty
to do in the garden. Pruned
the Buddleja near the south-eastern
corner of the house, but didn't have the energy for much more.
One of the results of the heavy winds at the beginning of the month was that we had a lot of
detached Kaffir Lime leaves. Did a
bit of investigation of possible dishes to cook with them, and finally came across a recipe
for opor ayam. The recipe itself seems to be of
dubious authenticity, and I made it worse by substituting various herbs and spices, but the
result wasn't too bad. About the worst thing is probably that there was too much coconut
milk—one of the disadvantages, I fear, of using canned coconut milk instead of the
old-style thick and thin coconut milk. I wonder what people in Indonesia do nowadays.
My leg was feeling really good yesterday, but today it was worse again. I suspect that for
some time to come it will need more rest than I have been giving it. Into town today to see
Heather, and once again felt a lot better after leaving—for a little while. I'll be
glad when this is all over.
Our Paulownia kawakami
(Sapphire Dragon) is about
2½ years old, and about 4 metres tall. This
year, for the first time, it looks set to bloom. The buds are about 1.5 cm wide, and
gradually they're lifting a corner to take a look out:
While taking the photos of
the Paulownia, considered
that it's time to switch over to the new operating system for dereel. What's holding
me up? I hate rebooting. dereel was only up 50 days, but still. On top of
that, I can see lots of issues with X. But something almost forced my hand. Came back and
found messages like:
That was only part of it. The entire disk subsystem (including other disks) appeared to be
in some kind of hang. Even the log messages didn't all make it there: there were
several FAILUREs as well. I had no option but to reboot (and no option what system
to boot: the new one isn't finished yet), pondering how to restore the disk: it's a 1 TB
file system, and I don't back all of it up. But it came back with no problems at all. Was
this really a disk error or something else masquerading as a disk error?
Spring is coming more slowly than usual, but it's coming. There's plenty of work to do, and
with my leg I can't do much of it, as my attempts a couple of days ago showed. So today CJ
came along to put up the fence around the vegetable patch to protect it against wind and
rabbits. To my surprise, that took up all day, and he still wasn't finished:
The Erysimum “Bowles mauve”
that we transplanted last year looked quite
unhappy at the start, but they have completely changed their mind. The following photos
were taken before the transplant, afterwards and (from the other direction) last weekend:
By contrast, the Erysimum
cheiri that I transplanted into the north bed over 2 years ago have barely managed to stay alive—until now. This year
they're finally starting to flower:
I've been using dircproxy to
work around the continual dropouts that SkyMesh, my ISP, blames on my modem. It works relatively well—I've been
connected for 3 months now—but it only allows one connection per nick. If I use a
proxy, I might as well be able to connect concurrently from different places (in particular
from the lounge room and the TV). Some months ago I installed bip, but couldn't get it to work. On connection I got the
message:
Error in protocol, closing...
What does that mean? It still happened today, so first upgraded to the latest version (from
version 0.7.2 to 0.8.6; how I mistrust software with a version number < 1), but with no
change. Did a bit of googling and discovered that this
is bips inimitable way of saying “I just got a SIGSEGV”.
Went looking for other alternatives—somebody mentioned quassel, but when I went to build it, it
required qt, with
which I have had bloat issues, so gave up on that pretty quickly.
Back to bip and experimented. Success! Well, workaround: it seems that bip
doesn't expect spaces in passwords, and its way of telling you is to SIGSEGV.
Clearly a bug; but why hasn't anybody fixed it? People claim that one advantage
of open source free software is that you can fix bugs yourself. And so you can,
if you can program. But I can't be bothered, and it seems that nobody else can be either:
the disadvantage of free software is that it exists to scratch an itch. No itch, no
improvement.
Everybody uses Google, of course, at least the
search engine. I use it all the time, and because of the long satellite delays, it's much
faster to set my default settings to return 100 replies. But lately that hasn't been
happening; I've been getting only 10. Changing it in the settings doesn't help: it tells me
it has saved the settings, but it lies.
That would be strange enough, but during my experimentation I discovered that it only
happens when the display language is English. If I set it to German or Dutch, I can get 100
replies—but of course, the selection changes and I get more German or Dutch replies at
the top, so that's not a solution. But what's causing it? Debugging this sort of thing
seems to be almost impossible. It's not even clear whether it's Google or the browser.
Discussed yesterday's problems with bip on
IRC today, and somebody came up with the possibility that the IRC protocol might not be
transmitting passwords with spaces correctly. Did some tracerouting and found that yes,
indeed, it was all arriving:
02:29:41.871880 IP sat-gw-ext.lemis.com.55016 > 1032.x.rootbsd.net.7778: P 154:233(79) ack 43 win 32768
E..w..@.).\F..p..V.....b..#g..*gP...>Z..PRIVMSG #bugs :E..>.4@.).....p..V.../.b..<J.<..P.......pass grog:FOO BLAH:oz
The payload here is grog:FOO BLAH:oz (user, password and internal channel ID
separated by :). Somebody else came up with RFC 1549, claiming that it didn't allow
spaces there (section 2.3.1):
<middle> ::= <Any *non-empty* sequence of octets not including SPACE
or NUL or CR or LF, the first of which may not be ':'>
<trailing> ::= <Any, possibly *empty*, sequence of octets not including
NUL or CR or LF>
In particular, <middle> doesn't allow spaces. But the way
I interpret this is that this is just the way IRC parses the message. Clearly you can use
as many spaces as you want in a message text. And in this case, it's bip that is
doing the parsing according to its rules. I still see it as two bugs: first, not allowing
spaces in passwords, and secondly (which nobody disputes) SIGSEGV is not the way to
report it.
Things are gradually coming to life as spring progresses. Last year I planted
a Cyclamen under one of the
birches, and it sat there doing nothing all through the spring; but finally it's coming to
life:
There's also one plant that I can't identify. We've seen it several times; Yvonne thought it might be
a Forget-me-not, but it's clearly
not that. Is it a plant or a weed?
The mystery plant is in the first photo. The plant in the middle of the second photo is
different: it's self-sown Oregano, which
is popping up all over the garden, along with self-sown strawberries (top left).
It's not clear whether the other one is a weed or not. I wonder if
it's Antirrhinum, which we planted in
the area last year. I could well imagine it self-seeding extensively.
One of the problems with long-haired dogs is that they get dirty. Today Nemo crowned himself in glory by rolling in mud at Chris' place,
and we had no option but to bathe him:
These are taken with my 300 mm Hanimex and a 3x teleconverter. They're really barely
acceptable. Yes, better than some I took with the toys I bought last year, but still not
nearly good enough. To get any kind of depth of field I needed to stop down to nominal f/11
(really f/33), and even then I couldn't get the whole field sharp enough. On top of that,
chromatic aberration made a real mess of things, and because of the aperture I had to choose
a really slow shutter speed, between 1/8s and 0.8s. The tripod kept the lens steady, but
the birds didn't keep still enough. I really don't know how to improve the situation.
For some reason my TV reception is deteriorating again. I can't work out why, but it seems
to relate to the new USB tuner, which for some reason is being detected as tuner 0, though
it's at the end of the antenna cable daisy chain (it has to, since it has no antenna out
connector). Rebooted the machine and things seemed better. I wish I knew why this stuff is
so flaky.
Another power failure in the middle of the
night, this one apparently very short: only the clock radio in the bedroom was affected.
Things do seem to have got better since the Nottles moved in across the road. A pity
they're moving out again already.
Sasha Ivkovic along today to bring back the map he borrowed from me last Friday. Did a bit of talking: he runs a class for second year
students every year, notionally to learn project management skills, but he prefers real-life
projects, so he encourages people to participate in free software projects. But which? It
seems that he spoke to rms about
it last week, but rms thought that it would just be a waste of the maintainer's time. I
don't think the FreeBSD project would see it
that way: we have a project
ideas page for exactly that purpose. Maybe we can get some contributions from them.
Of course, what's the benefit? The average second year student probably can't contribute
that much (maybe that's what rms was thinking), and how much can you learn of project
management by participating in any free software project? We'll see, hopefully.
It's been some time since I finished the software upgrade intended to become the
new dereel, and I've been treading water. First I need to get my head around
jails
The more I look at it, the more it seems that the tuner problems I'm having are mechanical
in nature. While playing around with cables, managed to wedge the computer a couple of
times, and after powering down and reseating everything, it seemed to work well again. The
only question is: how long?
Off to Stawell today to visit
the Spring Orchid Show, which
appears to be particularly famous. Cut across country, and the worst thing that could
happen did happen: my GPS navigator hung itself up again, so we didn't know very well
how to continue. I think I'm going to give up on the navigator: it's hardly worth trying to
put more maps on it when I can't even rely on it to run when I need it. I'll get ALDI to repair, replace or refund it. If they do the
former, I'll sell it or keep it as a backup.
In Stawell, had a lot of fun finding the place. “Entertainment Centre, Main
St.” was all the information we had, and it wasn't on the VicRoads country street
directory. After my GPS navigator came back to life, confirmed that it didn't know
about the entertainment centre either. There were also no signs. Finally ended up asking
somebody, and it seemed that we had been past it twice. They had put up signs—no less
than four next to each other directly outside the Entertainment Centre, but nothing further
away. Even directly in front of the venue there was almost nothing to be seen (first
photo). The one sign potentially visible from this viewpoint is hidden by the billboard on
the left. The only sign visible at all was at the rear of the building, carefully placed to
that you can't see it when driving a car (second photo), and where you can't easily access
the hall.
Getting out of the car, tried turning off the hung GPS navigator. It worked! I had
previously disconnected the USB cable (which charges it) to let the battery discharge, but
before trying to turn it off, I reconnected it in the hope of getting it to shut down on
disconnect (which is the default behaviour). Was that the reason I was able to turn it off?
On the whole, I think not: it just hung in a different way.
That's about 60% of the total exhibition, but then, we were in a small country town.
I was also not overly interested in the “artistic” arrangements; after some
consideration we came to the conclusion that the best one was the one on the left:
They had a vote for the best exhibition. I was going to withhold my vote, but they were
giving away small Dendrobium plants to people who voted, so I did after all:
As planned, and with the help of some of the people at the show, bought a couple of orchids:
A Cymbidium “Careless
heart” and a small Dendrobium
with a big name on its tag: Dendrobium (snowflake x merlin) x Hinazura “Winifred
Fortesque” [sic]”. The Cymbidium had a couple of flower spikes on it
which had come off, so now we have orchids in the vases for a month or two. The Dendrobium
looks pretty dried out, but it seems it's healthy enough.
It's interesting to compare the differences between here and Melbourne. There the orchids
cost in the order of $45; here we paid $25 for the Cymbidium and $12 for the Dendrobium.
The advice was also very different. It looks like we've opened a can of worms here.
It was still early when we let the show, so decided to head
for Halls Gap and take a look around
there. On the way noticed some strange weeds along with the
inevitable Capeweed (clearly visible in
the first photo):
They're Gazanias, of course, but growing
wild. They're also some of the prettiest I've seen, so did my part to help remove weeds and
took those four plants with us.
Had lunch in Halls Gap: “Souvlaki”, really a kind
of Gyros. I suppose I should have known
what I was in for when I had the choice of beef, chicken or vegetarian, but not lamb. The
weather was really mild: it must have been over 20°, something we haven't experienced at
home for months.
Once again to the visitor centre, where I once again got
very little information about what to see. In view of the condition of my leg, which is
healing but not yet good, decided to try a couple of places potentially reachable by car.
That appeared to be Mount William, so set off, on the way passing a turnoff to Silverband Falls, which aren't
marked on the official tourist map. But they required a 1 km walk, normally nothing, but
not appropriate for today. When we got to the turnoff for Mt. William we discovered that
the car park is 2 km from the summit, so decided against going any further. Potentially we
could have seen something from the car park, but it was 20 km there and back, and without
any useful documentation we decided against it.
Headed off back home, and in Pomonal saw
a sign pointing to a wildflower nursery. Headed off to take a look, and were very
pleasantly surprised: they had not just the nursery, but a garden full of particularly
unusual plants and shrubs:
So we bought one of those, and also
a Eucalyptus macrocarpa
(which they claim will only grow to 2 metres), and
a Leionema dentatum (“shrub 1 to
2m. Masses of cream flowers in spring”), and
a Orthrosanthus polystachus
(“Tufting grass-like foliage. Blue flowers in spring”). There's nothing in the
description to say so, but it seems that this last is a bog plant.
Unfortunately, they didn't have any seedlings, so they gave us some of the seed capsules.
Now to see if we can get any viable seeds out of them.
Back home: the temperature had only hit 17.2° here. I wonder if the Grampians are really
warmer, or whether it was just today. Planted the Gazanias, which had split up from 4 to 17
plantlets. I wonder how many will survive. While we were away, CJ had finished the
vegetable garden, as planned. Now we don't have much excuse for not planting things.
Saturday's house photo day, of course, and so had my usual work with that. But today there
was the backlog of yesterday's photos to do as well, with the result that it was nearly
15:00 before I was finished. Now I have only panorama photos of the garden, though in some
cases I wonder if it makes much difference. Here a single photo of two weeks ago, and
today's “panorama”:
This variation is also showing a weakness in my photo processing. At the moment I have a
list of which photos I should take, and I feed that into some scripts which ultimately
create the photos. But I've had up to 60 individual photos to make one panorama (one to 5
photos at different exposures to create an HDR base photo, and up to 12 base photos), and
manipulating them is difficult. I should think of a way to codify them in a more compact
manner.
I started counting hits on my diary about 10 years ago. At the time I was getting about 35
hits a day; lately the number has been hovering around 130, with more from the RSS feed.
And then today I made a mistake in my markup. It usually doesn't show on the web page, but
I have checks in my PHP code for that sort of thing, and they send me mail messages. But
how many! In 2½ hours I got 741 error mail messages. Off to investigate and discovered:
There are a surprising number of automated requests for things like:
That's nonsense, of course, for a number of reasons: firstly, those two requests are for
exactly the same page, just with one photo in different sizes (imagesizes
parameter). There were hundreds of references to this page in all different sizes. In
addition, the dirdate parameter shows that the real date was 16 August 2010, no longer in today's diary. I'll have to think of how to handle
that.
My scripts are out of date. When I started, I was using plain HTML, and there were no
parameters, so I didn't strip them. Those two hits above would show as two separate URLs.
That's easy enough to fix, and it made an enormous difference to the number of hits:
before it reported 126 hits for yesterday, and afterwards it was 6213. Of course, with
this kind of automatic reaping, it's not clear what use it all is.
Yvonne left early to take Nemo to a tracking class, and I had to make breakfast myself. But I wasn't hungry,
something that seldom happens to me. Was it the weather? I only ate half a breakfast, and
skipped lunch, but apart from that didn't feel in any way unwell. Was it due to the sudden
change in weather maybe?
For some months I've been taking photos of the flowers in the garden on the last Sunday of
each month. This month, strangely, there's not that much difference to be seen.
The Calendulas and tulips are now
blooming happily, and the very
first jasmine flower has opened:
About the most interesting things are
the Gazanias that we brought back
from Stawell on Friday. I pulled them out of the ground, left them to dry out in the
car for several hours, then just planted them in the ground and watered them in. And the
flowers on them are still flowering! Here “before” and “after”
Had a strange problem processing my photos today. One photo, and one photo only, was
corrupt. I use the Ashampoo photo optimizer
to enhance the images, and since the problem occurred after using that, I suspected that.
But the image looked strange to begin with. The sky is blotchy:
convert (one of these ambiguously named ImageMagick programs) got quite upset with the original, but not with the result. It
produced copious quantities of messages like:
convert: Corrupt JPEG data: bad Huffman code `_Ashampoo_Photo_Optimizer_Backup/Cathedral-3.jpeg'.
convert: Inconsistent progression sequence for component 0 coefficient 1 `_Ashampoo_Photo_Optimizer_Backup/Cathedral-3.jpeg'.
convert: Corrupt JPEG data: 58605 extraneous bytes before marker 0xd9 `_Ashampoo_Photo_Optimizer_Backup/Cathedral-3.jpeg'.
Did some investigation and discovered that the corruption occurred when convert
converted the original to a more highly compressed version. The Ashampoo optimizer seems to
have silently fixed some of the breakage. But how could that happen? The clue came when I
looked at the list of photos to convert. The file name Cathedral-3.jpeg showed up
twice:
But that shouldn't be a problem. To convert the files, I use the target convert in
my photo Makefile:
convert:
grep -v '^#' makejpeg | by 5 ${TOOLS}/imgconvert
The relevant part of imgconvert is:
if [ ! -s $NAME -a ! -s Components/$NAME ]; then
$ECHO convert -interlace line -quality $QUALITY -rotate $ROTATION $SRC $NAME
fi
Clearly there's a check for the existence of the file there, so it should only process the
first occurrence of the name. And that's what it used to do. But now I have a
multi-processor machine, and I do it in parallel (by 5, i.e. 5 processes at a
time), so these two conversions occur concurrently. And it seems that somehow two
different instances of convert write to the same file. I need to work out how that
could happen, but at least I've tracked down where the problem is occurring (and have since
fixed it).
Into town again to see Heather, who decided that my leg is improving enough that I don't
need to come back unless something goes wrong. As she predicted two weeks ago, it will take between 4 and 6 to heal, but things aren't too bad now.
It's just a nuisance that I still can't do much work in the garden.
While in town, to Bunnings to buy some
irrigation stuff. We're pretty much ready to plant the vegetables now, but we need water.
Now if I only had the agility to lay the stuff.
That's Piccola, of course. They just sat there for
some time looking at each other. Lee and Ray are moving out soon (they only bought the
house in March), and heading to
the Dominican Republic. I
wonder what will happen to the cats.
Today was my birthday, and for the first time I got lots of birthday greetings: 4 on IRC, 8
email messages, and 15 on Facebook. Half of the
email messages were from web forums which I have joined and to which I mistakenly gave my
correct date of birth.
Now I don't really use Facebook. I signed up years ago and haven't seen any reason to
leave, but about the only thing I do is go there every few months and attend to any
“friend” requests. And in previous years I haven't received this much
attention. But how do I reply? The correct way is on Facebook, of course, but that's such
a pain. And the messages say something like this:
Date: Tue, 28 Sep 2010 00:08:13 -0700
From: Facebook <notification+ysrrnwrr@facebookmail.com>
Subject: Lim Kim Yong posted on your Wall.
Lim wrote:
"Happy Birthday Greg"
Reply to this email to comment on this post.
So I replied, and got a message back saying:
Date: Tue, 28 Sep 2010 00:19:18 -0700
From: Facebook <notification+ysrrnwrr@facebookmail.com>
Subject: Your Facebook comment could not be submitted...
The comment you tried to submit via email could not be submitted because of the following error:
You must enter text before adding a comment.
Please try submitting this comment through the Facebook website instead.
What kind of message is that? No reference to which message, not even
an In-Reply-To: header, and how can you add a comment without entering text?
Somehow all this stuff seems completely broken. But it's amazing how many people I respect
use it. I wonder why.
The brief warm spell a couple of days ago was far too brief, and today the temperature
barely hit 13.5°; the average was 7.8°, and we had the air conditioners turned on all day.
That doesn't mean they were running: they have a mind of their own. For some time I've been
trying to find incontrovertible evidence of breakage in their temperature regulation
firmware, and there are at least two issues:
It does not set the fan speeds correctly. I have them set to HIGH all the time,
in the hope that that will blow the warm air as far from the head as possible and not
confuse the stupidly placed temperature sensor. According to the instructions, this
means that the fan will run at full speed at all times, even when it makes no sense.
But that's not what happens. The fans spin down to the two lower speeds, or even turn
off, when the temperature is nowhere near the set point. In the bedroom I set 30° and
can sometimes get it at 18°. This isn't a question of capacity. If the units don't
run, they won't reach their potential capacity.
In my office I set the temperature to 23°, which sometimes is more or less correct.
Again I have the fan on HIGH. In the evening, I turn the temperature down to the
ridiculously high minimum of 16°. And very often, the fan will speed up! This happens
on other heads as well, and on the other external unit, so it's not a fault with one
individual component.
I have many other complaints about this
system, but I think these two are such a clear indication of broken firmware or design that
nobody can deny them.
Over the last few months I've commented on my ALDI GPS navigation system,
in particular how it continually hangs, and how I haven't found a way to reset it when it
does. There's a hole in the front of the unit just above the charge light on bottom left,
but inserting a pin doesn't do anything interesting:
Today, almost by coincidence and while looking for a replacement, I found the correct
hole. It's on the side (where the paper clip is inserted), and it's even marked, in an
almost completely illegible impression on the case:
I'm quite proud of being able to get a photo of it: it's even more difficult to see with the
naked eye. It effectively powers off the unit, and you need to power it on again;
presumably it just interrupts the battery supply, which makes sense. So now I need to
reconsider whether I should buy a new one.
A couple of days ago I discovered that I was
getting lots of diary hits for long-expired entries. Finally got round to looking at my
photo code, and it's ugly. Managed to remove some of the parameters, with the side-effect
that the URLs have got a lot smaller. Now if I get a URL like diary.php?dirdate=20071110&image=dam-nw.jpeg&imagesizes=302213, I redirect to the
diary entry for 10 November 2007. What I can't do is show the photo:
firstly, the image name is very often just plain wrong (there is no reference to that image
in that month's diary; possibly something I should investigate), and secondly it wasn't used
properly anyway. But at least it doesn't give null references on the current diary page.
The other issue is that I shouldn't be letting the bots index my diary. That's simple:
<meta name="robots" content="noindex" />
But instead they should index my monthly diary (currently diary-sep2010.php). How do I get them to do that? They don't seem to do it by default, possibly
because they don't find anything they haven't already found in the current diary page.
According to the documentation, the meta tag content shouldn't stop the crawlers
following links. But that's something I'll have to observe.
Getting the photos was a nightmare. The light was from the back, so I needed flash, and
once again the exposure was terrible. Took the first shots with the camera's flash, and
wasn't happy with them. Got my Mecablitz 58 AF-1 O digital flash and tried again—and things were much worse!
The first of the following photos was taken with TTL automatic exposure, and it's
underexposed by at least 3 EV. Even with +3 EV compensation (the maximum that Olympus allows), it was barely enough.
Sometimes I wonder why I bother with automatic exposure.
The images above are as I processed them at the time.
Ten years later I reprocessed the photos with my standard
workflow, which showed considerable improvement, for example:
CJ along today to finish off the work in the garden, but Yvonne hijacked him for a couple of hours to start burning off the fallen branches we've
acquired over the winter and to do some trimming work that the Shire Council should have done:
I wasn't mobile enough to help, but Yvonne did some work with him. After finishing off the
garden fence (but not the gate, which required rivets I don't have here), they put the
baseboards on the verandah, the final and somewhat
belated touch:
We've decided to transplant our citrus trees, all of which look quite unhappy. If we put
them in the south of the eastern part of the garden, in front of the garage, and put a wind
break behind, we hope that they'll do better. That requires putting rails over the posts
and hanging mesh down, like we've done in the north-east. Found some old timber from the
verandah construction, which we had rejected because it was too twisted. I hope that it
might straighten out, but at the moment it doesn't look too convincing:
This horrible SkyMesh satellite networking
service is really bad again. It comes and goes in spells, and there seems to be little I
can do about it. Sometimes I can go a couple of weeks without a dropout, but today I had 7.
Paul Rees, my contact person at SkyMesh, who may or may not be the general manager, has been
successful in deflecting my complaints by claiming that the modem is defective. I don't
believe this, but I can't prove my viewpoint any more than he can. But he has one ace in
his hand:
Based on IPSTAR\222s expertise as the satellite operator and our support staff\222s
experience in troubleshooting problems like this, we believe it\222s the modem. What do
you believe it to be?
The \222s are the result of his use of invalid (probably Microsoft) characters in an ISO
8859-1 text. If I leave the originals, they will render incorrectly and invalidate this
page. His justification for this claim is not from any diagnostics (they don't do no
steenking diagnostics), but his interpretation of IPStar'sreport on the matter some
time in early June:
Signals all look to be very good so ODU would be doubtful of an issue but when the mass
dropouts were occurring we can see failed login attempts (indicating Tx (transmit) issue)
and quite a lot of 'PowerUp' as the logoff cause. As the Tx side of the UT is powered
along with these logoff causes I made my recommendation in TT:24061 this morning at 9:21am
AEST to check power onsite with the CPE (Customer Premises Equipment). If site is found to
be fine then the next point of call would be to look at the modem as it would most likely
be faulty in this case. Just to finish off, there was no service issue for this customer
relating to equipment with the IPSTAR gateway or SkyMesh service.
Nearly all the “PowerUp” events were manual: I power cycled the modem in the
hope of unwedging the link, usually in vain. Since then, most of the issues we've seen
haven't been of this nature, as I tried at length to explain to Paul, but either he doesn't
understand, or he chooses not to. Most of the issues we have are a generic event “No
response from modem”, clearly something that could have multiple causes, and he made
these claims appending a log showing
multiple “No responses” and one “Modem powered down”, which I did
manually. I can't see any connection
with what the IPStar people described.
OK, there's an obvious way to check his claims: install a new modem. But he wants me
to pay for it up front, a sum of $665.80 for a refurbished modem. Why up front?
We're not prepared to pay Skybridge to send you a modem because we have no way to get it
back from you if you decide to keep it
That's just plain insulting. In addition, he has a direct debit authority for my account,
so it's also incorrect. He suggested that I should buy a second-hand modem on eBay; what good is that? If I buy one and things don't
improve, then he can say that this modem, too, is defective. Some of the responses have
been decidedly childish:
>> . You don't want to pay for an exchange modem unless you are sure
>> that it will solve the dropouts
>
> Would you?
Yes, I would.
I'm convinced that this is an attempt to shut me up, and so far it has worked. But I'm
quite sure that this problem isn't with the modem; all the indications are that it is in
their network, including:
The dropouts usually occur in conjunction with various other problems, such as
congestion. That was the case today, as the excerpts from my satellite monitoring logs show:
In this particular case, I didn't record a dropout, since no single entry showed
complete lack of connectivity. But this went on all morning, and it suggests to me that
the dropouts that did occur weren't related to the modem.
In many cases, I don't lose BST connectivity. This indicates that the link
between the modem and the ground station is still functional. I can often ping the base
station but not beyond. Again, Paul doesn't want to understand this.
The real problem, though, is: what do I do? If I go anywhere to complain, such as the
Department of Broadband, Communications and the
Digital Economy, the people I will interface with will have even less understanding of
the issues than Paul, and without some really conclusive proofs, they'll agree with him.
And these toy modems seem to provide no way to debug the issue. I'm sure, though, that if
SkyMesh had the goodwill to examine the problems more easily, they'd find a number of issues
not related to satellite links.
So, as Paul asked:
What other problems would you have us fix, that don\222t affect all
our other customers, just you?
There are two obvious answers there: first, I can't confirm that no other customers have
complained. But even if they haven't, that doesn't mean that they're not affected.
Years ago I had to point out to Chris Yeardley that her
satellite connection wasn't working correctly. What was the problem? Continual dropouts,
of course. And she's technically savvy. Most users wouldn't understand enough to localize
the problem.
While talking about Facebook, thought about
how long it's been since I used Orkut, the first
of these social networking sites, which some people I know have never heard of. Looked in
again and found:
Not many new features. In fact, of course, it's probably flash breakage of another kind.
More bizarrely, though, Rob Levin is
still there, though he died four years ago. I wonder how
many other corpses are still on the web.
More playing around with CSS and my diary
layout today. The big issue I have is to put normal text in a limited width box (about 70
characters) while using the full width of the window for photos and computer output. The
CSS model doesn't really think that way: it's a strictly hierarchical arrangement, so each
block must fit inside the containing element. To get what I want, I'd have to completely
rearrange the way I see a document, placing arbitrary chunks of text into a
<div> of appropriate width. That's ugly. It also means
going back and modifying 158 diary files by hand. What I need is a way to do it with as
little overhead as possible. So far, that still hasn't worked, but my approach has been to
limit the size of paragraphs and <li> tags. And there's
the problem: limit the width of the <li> and you limit the
width of everything inside it. If I don't limit it, many of my old pages will render list
elements in full width. So I need a limitation with exceptions, certainly not the way I
would have done it if I had started out with this knowledge. But gradually I'm getting
there.
In previous years we've had Ashley Musgrove come to spray the paddocks for weeds, but this
year we've been unable to contact him. Despite the weather, the year's getting on, and we
had to get it done, so today Don Fortescue came along and did things pretty quickly:
To the immediate north of the house we have a verandah, then an area of gravel, and then a
bed. It's been a pain to keep the gravel free of weeds, and we've been gradually planting
bulbs that don't need watering beyond what the rain brings (in the winter). We have also
had self-sowing of cat mint (which leaves the cats cold), thyme and violets. I suspect the
last will die off in the summer, but we'll see. Today Yvonne planted some more plants: Sedum “gold
mound”, some Carpobrotus and
some Watsonias:
My external mail server has been up for two years:
Thu Sep 30 00:55:32 UTC 2010
12:55AM up 731 days, 2:32, 3 users, load averages: 0.06, 0.01, 0.00
That's the longest up time I've ever had. People point out that this leaves it open to
vulnerabilities. They mean “more open”, of course, and yes, that's true. But
how much so? Since my Tandem days up time has been (almost) everything, and I don't see why
I should worry about vulnerabilities on a site that is not very interesting to an attacker
in the first place. If I can keep it up another year, I will.
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.
