First day of the conference today, and the first time for some time that I've been here almost purely as a participant. This year seems to be particularly good; we've had the highest attendance for five years, and everything seem to be going smoothly.

Started off with Senator Kate Lundy, shadow IT minister (and possibly the real one in a couple of months; today was also the start of the election campaign). I heard her at the “Open Source in Government” conference in January, but this one was a different talk, and much better.

Then listened to a few talks on system configuration. Adrian Close's talk on “Kara” was particularly interesting, since it's a different approach to the system configuration scripts I've been working on since the beginning of July. I suppose I'm going to have to look at that in more detail. Then Michael Paddon, ostensibly talking about scalable remote firewalls—an interesting idea in themselves—but in fact more a discussion of R-trees, which might be an interesting thing to look at in context of Monkey. He later gave me a link to some which I must look at, though he says the code is not functional in that form.

After lunch heard more about VoIP from Andrew Rutherford; I still need to look at that in more detail. Steve Bellovin did an interesting and thought-provoking talk on identity and privacy, probably the closest we came to the conference theme “Who are you?”

The Annual General Meeting went fairly painlessly, though it's clear that we still have two different groups of opinion about what to do with AUUGN. Frank Crawford's prepared to get involved, which is a great idea.

Networking reception in the evening, then out to the Indian Restaurant around the corner, which was better than I recall it, though it's always a bit of a waste after the snacks at the reception. Relatively early to bed.

Interesting talk this morning by Theo de Raadt about the work they've been doing in OpenBSD to make buffer overflow exploits much more difficult. I had heard of much of it before, but it was well presented and made a lot of sense. Later in the morning Warren Toomey talked about what they're doing now with TiVo. It's interesting to think how much has happened since our first discussions of this topic. He and Luke Mewburn showed me a few things that I didn't know. Overall, though, I still don't like TiVo too much, and much of the demonstrations just went to confirm my prejudices.

In the afternoon I was session chair for a keynote speech by Jess Healy, the Democrats lead senate candidate for Victoria at only 19 years old. She gave a good talk within the time constraints allocated. After that Bdale Garbee with issues of trust in the Debian project; good food for thought for the BSD projects too.

In the evening the conference dinner with its usual madness. We got started earlier than usual, even before the entree, and we were decidedly flagging by the end. David Purdue had organized a wine tasting competition, unfortunately with a 1995 Lenswood Riesling which was so far beyond its prime that all the “experts” thought it was very young. David Newall won, or, as he put it, was the last one left sitting.

Up later this morning, as usual after the dinner last night, missing Steve Alford in the process. It's a pity they put good talks on on Friday mornings.

Most of the good papers seemed to have been in the first two days, and I didn't go to anything of particular interest; instead spent a lot of time in the secretariat arranging photos and taking more for people who wanted some. I've decided that this Nikon Coolpix 880 is not the best in the world for flesh tones, and it often leaves people looking as if they have been drinking all night (today was, admittedly, not a good day to test that).

The conference wound down rather suddenly, as AUUG conferences tend to do, though we spent a bit of time in the bar talking about all sorts of futures. Pia Smith, president of Linux Australia, had turned up rather suddenly in the middle of the night, and we spent a bit of time talking. Discussions with other people showed a big discrepancy between the way I think and the way others think: most consider my discussions of AUUG's future and direction to be indiscreet, and since I don't want to upset them, I have reluctantly decided not to discuss them any more in this diary.

Then off with David Newall, Ollivier and Élodie to the Indian restaurant we had visited on Wednesday, this time with a little more appetite. David was telling me that CSIRAC, Australia's first computer, is now on display in the Melbourne museum, and that it's the oldest surviving computer in the world. Changed my return plans to see that.

Intended to go to the museum with Ollivier and Élodie this morning, but their time schedules didn't mesh too well (I keep forgetting how much work a baby is), so off by myself to the museum in Russell Street. A good thing that they didn't come with me: the museum moved to Carlton (just North of the Exhibition buildings, and not too far from where I lived from 1957 to early 1959, even closer to the hospital that treated me for the strange accidents I had with chemicals at that time) years ago, and I had to come back and get the car.

The new museum reminds me in concept of the pyramid of the Louvre: it's right next to a staid old building, and it's ultra-modern. Both museum and Exhibition building don't match up to their Parisian equivalents, of course. The CSIRAC exhibition was interesting for a number of reasons: I had never seen a mercury delay line before:

Just the presence of the machine was imposing:

Unfortunately from my perspective, it was aimed towards predominantly non-technical people. The drum was labelled “disk drive”, and though the machine was not running (understandably; maintenance and power would have been a nightmare), the display lamps had been wired up to a random on/off generator of about 8 Hz. There was almost no information about the technical details of the machine (I later discovered it was a 20 bit machine running at about 1 KIPS (0.001 MIPS), and that it had about 2,048 words of memory). I should have a chat with the people who did the exhibition; I consider that this is the world's most important old computer exhibition, and I could imagine a lot of interest from technical people world-wide. Also, it's possible that we could put some of Australia's early UNIX history in there too; this needs discussion before I can give details.

Revisiting this 11 years later, I discover I completely forgot to mention the other hardware there, parts of the Control Data 3200 that I had used (at a distance) at the CSIRO site in Clayton in August 1971. There was a memory cabinet and a tape drive:

Image title: CDC 3200 1          Dimensions:          1536 x 2048, 1252 kB Make a single page with this image Hide this image Make this image a thumbnail Make thumbnails of all images on this page Make this image small again Display small version of all images on this page All images taken on Saturday, 4 September 2004, thumbnails          All images taken on Saturday, 4 September 2004, small Diary entry for Saturday, 4 September 2004 Complete exposure details

 

Image title: CDC 3200 2          Dimensions:          2048 x 1536, 1032 kB Make a single page with this image Hide this image Make this image a thumbnail Make thumbnails of all images on this page Make this image small again Display small version of all images on this page All images taken on Saturday, 4 September 2004, thumbnails          All images taken on Saturday, 4 September 2004, small Diary entry for Saturday, 4 September 2004 Complete exposure details

 

After that, off home via Parkville, where I lived until about 50 years ago, and took a couple of photos of the house we lived in there, at 123 Park Drive. Not surprisingly, I can hardly remember it, and I don't think I've seen it for at least 40 years.

Getting out of Melbourne and the initial parts of the Western highway are a nightmare. The speed limit is 100 km/h or less almost everywhere, and there are continual signs warning about going to sleep at the wheel. One kept intruding into the corner of my field of vision as “ONLY SLEEP CAUSES FATIGUE”. On closer examination, that was “cures”, not “causes”, but it kept coming back. Despite a good night's sleep, I did indeed feel tired, and it wasn't until after Nhill (without any sleep) that I felt less tired. I suspect that slow speeds do, in fact, tire drivers, and that the higher speeds in South Australia make it easier to keep awake. I have a serious problem with Australian attitudes to road safety: it would be perfectly possible for me to be run in by the police in Victoria on the open, empty roads for doing 110 km/h (exceeding the speed limit), and in Germany on their crowded Autobahnen at the same speed (obstructing traffic). Where's the relation in either of those to absolute safety?

Quiet day, spent getting back to “normal”, though I didn't try too hard.

The repair of the Digitrex DVD player was too good to hold. In the ten days after getting the machine back, it didn't hang at all, a very long time by comparison with its prior behaviour. Today it did, however—exactly the same symptoms as before. In combination with the service report, I suppose that this means that the machine is not repairable, and that this problem is a known (and evidently irreparable) fault. I shall contact Digitrex and ask for them to refund my money.

zaphod is still crashing regularly every night—in fact, twice a night, as Yvonne discovered. Debugging isn't helped by various bugs in gdb that stop me from finding the name of the file: it's passed to namei as a pointer into userland, and the new version of gdb in -CURRENT refuses to descend into userland. The development version gdb6 simply crashes trying to dereference it. Looked at it with ddb and discovered a relative pathname (sample3.bz2.uu, FWIW; there must be about 30 of them on the disk) and no associated directory name, which is puzzling. Looks like more work with firewire.

Apart from that, spent some time catching up and investigating userland exits to implement file systems. Linux has a “loopback file system”, but it seems that there's nothing equivalent in the BSDs. Looks like another diversion ahead.

Another day spent mostly in catchup mode. Managed to get my mail down to more manageable proportions, but wasn't very successful in finding useful userland file system stuff. There are a surprising number of strange file systems in /usr/src/sys/fs/. Things like nullfs and portalfs seemed to be possibilities, but they're not really what I'm looking for. Didn't find anything very useful on the web either. It's amazing how much time you can spend looking.

It's looking more and more like I need to create a VFS loopback layer, not something I'm looking forward to. I need to learn more about the structures, so what better way than to kill three birds with one stone? zaphod has been panicking daily for nearly two weeks now with the panic message cleaned vnode isn't. Spent most of the day investigating that and documenting the paths through the VFS layer, which will also be useful both for my understanding of the code and also for the next iteration of my kernel debugging tutorial. As was to be expected, I ended up in a number of dead ends, but that's life, and it's probably instructive to document how I got there and why. As a result, didn't find the bug: in particular, it looks as if VBAD is the correct type for an unused vnode.

There's another project coming up at work, and I've been asked to review. I've been putting the matter off for nearly a month now, so it was time to do something about it. Today was the day, and it kept me busy nearly all day.

Also looking at the zaphod panic issue, which is still puzzling me, not helped by problems getting firewire debugging to work. I wonder if the software is too old; since it's development software, there's no particular requirement for interoperability.

In the evening, Seren finally gave birth to her foal, now two weeks overdue. It's a female, and we've decided to call her “Wantadilla Serenade”. It's been raining quite a bit lately, and the poor thing was born into a puddle of mud and couldn't get up without help:

Still, she seems to be healthy enough, and very big.

Some days I seem to get nothing done, and today was one of them. It's not that I didn't do anything; I worked all day long. But at the end of the day, there was nothing to show for it, except maybe a few loose ends tidied up. I suppose it comes back to the question of mail: “reading mail” is only part of the equation: a large proportion of my work goes into mail, but I look on it as mail, not work. Time to reconsider things.

Didn't do much today, either. Yvonne is thinking about getting rid of Seren, who hasn't been rideable for a long time, and giving her to a breeder. Today off to the other side of Two Wells to visit somebody; it looks as if we'll have a couple of horses less in the near future, but Serenade will come back after she's weaned (probably in early February, when we should have two other foals).

Yet another quiet day; finally went riding again, this time in the Pocock Road section of Kuitpo Forest with Yvonne and Michael Hickinbotham. Darah was in fine form, far from the way she was last time, and I ended up leaving Yvonne and Michael behind and going round the forest at high speed; at least we went a little longer this time, though I probably didn't get more than 10 km.

In the afternoon, more investigating the zaphod panics, which are getting more and more confusing as I investigate them. This may end up being a race condition and not a hardware problem after all.

More work on my crashing zaphod today. This is really quite a fascinating panic, and though I suspect I could fix it faster if I didn't document what I'm doing, I think that it's an excellent example of a kernel bug. Since the system only crashes at night and doesn't seem to do itself any harm in the process, there doesn't seem to be any priority in fixing the problem.

Conference call in the afternoon. Looks like I'll be implementing (or finding) a userland file system callout.

Gradually I'm making some headway about userland file systems, paradoxically after replying to a message about Vinum in a DragonFlyBSD mailing list. It seems that Matt is very keen on the idea, and is even thinking of migrating less important file systems, such as cd9660, to userland—once the interface works, of course. Also heard that ARLA does this, and that it's very portable. Unfortunately, the site was down, but there's still plenty to chew on.

The site is no longer down: it's gone. And I can't find anything on the web that even states what ARLA is.

You'd think that beer brewing was an established enough tradition that there wouldn't be anything very new to report, but it seems that enzymes work differently in German speaking countries and English speaking countries: the recommended optima for α-amylase and β-amylase during mashing are different in the two countries:

Sent a message out asking about the discrepancies. The answers should be interesting.

I didn't get many answers, but John Palmer sent a document describing the process in much more detail. It's not exactly a predigested answer, unfortunately.

More investigation of the file system layer today. The ARLA site was back up, and I spent some time looking at the code, which of course doesn't compile. sigh I suppose it's worth the trouble to fix things first.

Into town today to interview somebody for a job at Rocksoft, then down the other end of North Terrace to meet John Sanders, the new executive director of the IT Council of South Australia. Interesting person; he certainly has a lot of energy, and I can see a number of changes coming.

Digitrex has finally approved a refund for my abortive purchase of the GKX-9000 DVD recorder. That's mainly good news, but it means that I need to find some alternative. While in town, purchased a DVD+RW burner, a couple of tuner cards and a TV out card, then home to try them out and discover that an old version of FreeBSD -CURRENT recognizes one of the tuner cards and the TV out card out of the box. That's more than I expected, but I still fear that I'm going to have a lot of work ahead of me.

More work on ARLA today: looks like I have just become maintainer, though it looks as if there's a non-committer out there who really uses it. More administrative stuff.

In the evening, put together the TV equipment I bought yesterday. Given the fact that it all comes with software for Microsoft, it seemed reasonable to use one of the Microsoft licenses that I have had forced on me to try it like that first. How I hate Microsoft! It didn't work properly. The DVD playing software I have only works if the display card is set to output to monitor only; as soon as I set it to dual head (including TV), it hangs hard and needs to be restarted. What kind of use is that? Also, once again everything is done with toy “skins” like the ones I commented on a couple of months ago. To stop the program, for example, you need to find and click on the button marked POWER on the toy display. Even the skins themselves suggest some teenager having fun with developing his own fantasy boxes rather than anything you'd ever see in a HiFi shop. Am I the only person thoroughly put off by these toys?

Spent most of the day investigating the issues surrounding computer TV. They took a surprisingly long time, not the least because of the old issues with installing new versions of FreeBSD. Still, TV out works, for a minimal definition of “work”. It seems that the output is made by dithering the image, and the resultant quality is pretty terrible. Hopefully—when I get that far—an X display will be much cleaner.

In the evening, Marti Winn from Winnfarms came to visit. She's spending a few days looking at horses in the area.

Computer Video Recorder, take 1.

Started building a CVR machine. Basic hardware and software are:

• Athlon 1700 XP+ CPU
• 512 MB RAM
• 20 GB IDE drive
• NVidia GeForce 4000 based card with TV out:

  none4@pci1:0:0: class=0x030000 card=0x93631462 chip=0x018510de rev=0xa4 hdr=0x00
      vendor   = 'NVIDIA Corporation'
      device   = 'NV18.6? GeForce4 MX 4000'
      class    = display
      subclass = VGA
            

  This card has some way of storing its configuration on the card, and as it's currently set it outputs the same data to both ports. The Microsoft-based drivers call this “clone” mode, but it's clear from the quality that it's doing some kind of not overly successful dithering.

• An el cheapo Brooktree 878 tuner card:

  none0@pci0:6:0: class=0x040000 card=0x00000000 chip=0x036e109e rev=0x11 hdr=0x00
      vendor   = 'Brooktree Corporation'
      device   = 'Bt878 MediaStream Controller'
      class    = multimedia
      subclass = video
  none1@pci0:6:1: class=0x048000 card=0x00000000 chip=0x0878109e rev=0x11 hdr=0x00
      vendor   = 'Brooktree Corporation'
      device   = 'Bt878/Fusion878A Video Capture (Audio Section)'
      class    = multimedia
            

• A 3Com Ethernet card. The motherboard has an onboard NIC:

  vr0@pci0:18:0:  class=0x020000 card=0x764c1462 chip=0x30651106 rev=0x74 hdr=0x00
      vendor   = 'VIA Technologies Inc'
      device   = 'VT6102 Rhine II PCI Fast Ethernet Controller'
      class    = network
      subclass = ethernet
            

  Unfortunately, the driver support for this NIC under FreeBSD 6-CURRENT seems dubious; it gave me my first-ever experience of a kernel debugger backtrace on the home TV.

  So I'm using this card instead:

  xl0@pci0:12:0:  class=0x020000 card=0x00000000 chip=0x905010b7 rev=0x00 hdr=0x00
      vendor   = '3COM Corp, Networking Division'
      device   = '3C905-TX Fast Etherlink XL PCI 10/100'
      class    = network
      subclass = ethernet
            

• The on-board sound hardware:

  none3@pci0:17:5:        class=0x040100 card=0x76401462 chip=0x30591106 rev=0x50 hdr=0x00
      vendor   = 'VIA Technologies Inc'
      device   = 'VT8233/33A/8235/8237 AC97 Enhanced Audio Controller'
      class    = multimedia
      subclass = audio
          

Which software?

bktr_mem: memory holder loaded
bktr0: <BrookTree 878> mem 0xdddfe000-0xdddfefff irq 17 at device 6.0 on pci0
bktr0: [GIANT-LOCKED]
bktr0: Unable to allocate 1310720 bytes of memory.
bktr0: Unable to allocate 3555328 bytes of memory.
bktr0: Card has no configuration EEPROM.  Cannot determine card make.
bktr0: Pinnacle/Miro TV, Temic PAL I tuner.
    

The Xorg server recognizes the card without problems and builds a configuration file that shows only one output. Some others, noticeably Daniel O'Connor, have reported success with the Nvidia drivers, so I installed them (/usr/ports/x11/nvidia-driver). After installation, Xorg generates exactly the same configuration file as well. Reading the instructions (notably README and README.Linux) helped, and rather to my surprise, I was actually able to get it to run two displays under X after applying the following changes:

--- xorg.conf.orig      Sun Sep 19 20:48:58 2004
+++ xorg.conf   Sun Sep 19 21:27:57 2004
@@ -1,6 +1,7 @@
 Section "ServerLayout"
        Identifier     "X.org Configured"
-       Screen      0  "Screen0" 0 0
+       Screen      0  "Screen0"
+       Screen      1  "Screen1" rightOf "Screen0"
        InputDevice    "Mouse0" "CorePointer"
        InputDevice    "Keyboard0" "CoreKeyboard"
 EndSection
@@ -42,11 +43,21 @@
 EndSection

 Section "Monitor"
-       Identifier   "Monitor0"
+       Identifier   "Alfaskop"
        VendorName   "Monitor Vendor"
        ModelName    "Monitor Model"
+       HorizSync    31.0 - 36.0
+       VertRefresh  50.0 - 160.0
 EndSection

+Section "Monitor"
+       Identifier   "TV"
+       VendorName   "Monitor Vendor"
+       ModelName    "Monitor Model"
+EndSection
+
+# Primary monitor for, well, monitoring.
+# Connect a low-res monitor such as the ancient alfaskop to this one.
 Section "Device"
         ### Available Driver options are:-
         ### Values: <i>: integer, <f>: float, <bool>: "True"/"False",
@@ -64,7 +75,7 @@
         #Option     "NvAGP"                    # <i>
         #Option     "IgnoreEDID"               # [<bool>]
         #Option     "NoDDC"                    # [<bool>]
-        #Option     "ConnectedMonitor"         # <str>
+        # XXX Option     "ConnectedMonitor"  "CRT, TV"         # <str>
         #Option     "ConnectedMonitors"        # <str>
         #Option     "TVStandard"               # <str>
         #Option     "TVOutFormat"              # <str>
@@ -105,17 +116,29 @@
         #Option     "XvMCUsesTextures"         # [<bool>]
         #Option     "HorizSync"                # <str>
         #Option     "VertRefresh"              # <str>
-       Identifier  "Card0"
+       Identifier  "Control"
        Driver      "nvidia"
        VendorName  "nVidia Corporation"
        BoardName   "Unknown Board"
        BusID       "PCI:1:0:0"
+       Screen          0
+EndSection
+
+# Second device: TV
+Section "Device"
+       Identifier  "Video
+       Driver      "nvidia"
+       VendorName  "nVidia Corporation"
+       BoardName   "Unknown Board"
+       BusID       "PCI:1:0:0"
+       Screen          1
 EndSection

 Section "Screen"
        Identifier "Screen0"
-       Device     "Card0"
-       Monitor    "Monitor0"
+       Device     "Control"
+       Monitor    "Alfaskop"
+       DefaultDepth 24
        SubSection "Display"
                Viewport   0 0
                Depth     1
@@ -139,6 +162,20 @@
        SubSection "Display"
                Viewport   0 0
                Depth     24
+               Modes   "640x480" "800x600"
        EndSubSection
 EndSection

+Section "Screen"
+       Identifier      "Screen1"
+       DefaultDepth 24
+       Device     "Video"
+       Monitor    "TV"
+       Option "ConnectedMonitor" "TV"
+       Option "TVStandard" "PAL-B"
+       Option "TVOverScan" "1.0"
+       SubSection "Display"
+               Depth     24
+               Modes   "800x600"
+       EndSubSection
+EndSection
    

Later installed mplayer on teevee. Installation went OK, but getting it to run is a different matter. As in the past, I had trouble. Display across the network is impossibly broken, and for some reason I couldn't get it to work at all on the local (monitor) display :0.0. The message was:

SwScaler: using unscaled Planar YV12 -> BGR 8-bit special converter
BGR not supported, please contact the developers
FATAL: Cannot initialize video driver.
FATAL: Could not initialize video filters (-vf) or video output (-vo).
      

After a fair amount of discussion on IRC, and with help from google, found out the problem: this is mplayer's inimitable way of saying “I can't use an 8 bit display”. After setting the default depth for the display to 24 bits, it worked fine. What a pain! The diffs above include setting the default bit depth (for Screen0).

More work on the video recorder today. What a pain! Installing MythTV was complicated by some of the strangest code I've seen in a while:

 /* ci must be 0 */
 if( bitoff == 0 ) {
    (uint8_t)strm[co]= bitten;
    co++;
 }
 goto BAUCHWEH;

HIRNWEH:

 (uint8_t)strm[co]= bitten;
 co++;

The (uint8_t) casts (of which there are several) are illegal, of course, but the comments seem appropriate.

I think I was referring to the labels.

Fixed that, and at least got it to compile.

Investigating the Nvidia display board I had proved to be useful: rather to my surprise, I was actually able to get it to run two displays under X.

My father was taken into hospital in England yesterday with some strange problem which wasn't immediately apparent; he'll need examination on Monday. This afternoon tried to call him and got fully two minutes of nonsense explaining that I would have to pay up to £ 0.50 per minute to speak to him, round 10 times as much as it costs me to make an international call. After that, presumably because they couldn't find a way to charge me, I got a request to leave my name, after which they hung up on me. I'm really astounded that they would do this to their patients. They also don't allow mobile phones in the hospital, ostensibly because they interfere with their electronic equipment. I strongly doubt that; it looks more like money-making to me. I certainly hope that their medical treatment is better.

More work on the ARLA stuff today. I got a message from one of the people involved in the FreeBSD port, which is currently broken as the result of recent changes in the FreeBSD bio layer. So what do I do? I'm not married to FreeBSD. It's supposed to run on NetBSD and Linux as well. Considered the options and decided that running it on NetBSD would have some advantages (Linux is too different at this stage, and the NetBSD port would help me fix the FreeBSD port), so downloaded an ISO and started installing it. Too late, I realised I had installed it on a SCSI drive whose controller has no boot ROM, so I won't be able to boot from it. Sigh.

Also reading about AFS, which I've never really looked at before. It's quite interesting, though performance doesn't seem to be its forte.

Still can't get through to my father: it really looks as if they're rejecting my calls because they can't reverse charge me. Sent a message to the webmaster (only address I could find on their web site), asking for the message to be forwarded to the correct people. Atypically, received a reply from the webmaster within an hour giving further information, copied to the right people, and wishing my father a speedy recovery. When you have this kind of experience, it's easy to look on the whole institution as bad, but evidently that's not the case here.

Still working on getting test equipment set up today. For some reason, a NetBSD i386 CD-R of the latest release (1.6.2) doesn't install cleanly: it seems to contain the wrong kind of binaries. Spent a lot of time installing over the network (the first time I've ever done that), and then building the rest locally. I wish this didn't take so much time.

Finally got through to my father on another phone line. I'm beginning to wonder whether it's not just a very poor user interface to a voice mail system.

Federal elections are coming up in Australia (on 9 October). We've heard the slanging matches in the media (summary: “Who's the bigger liar?”), but I thought that was probably not a good basis for a vote, so I spent some time investigating the web sites of the parties. The results weren't very helpful:

• I can't find anything on http://www.liberal.org.au/. They have a link “Election Policies” which looks promising, but it's a collection of large PDFs, totalling hundreds of pages and with misleading titles like “Supporting the Territory” (it turns out they mean the Northern Territory; though this is an accepted abbreviation, in this context it suggests that they don't know that Australia has other territories, nor indeed that it is itself a territory). The Liberals also expect me to use specific software, which I don't have, so I suppose I'm not welcome anyway.

• http://www.labor.org.au/ doesn't seem to be interested in the election. Apart from using it as a way to slate the Liberals (“Liberal's IR Lies”). the only policy they mention (indirectly) is the “Indigenous Policies”. The word “election” does not appear on their home page.

• http://campaign.democrats.org.au/index.asp get it almost right. They have a link to http://campaign.democrats.org.au/minifesto.html at the top of the front page, and apart from being somewhat unsure of itself (“Please take the time to consider this short version of the Australian Democrats? election platform”) gives the kind of overview (not necessarily the content) that I was looking for. The Democrats are also the only party I've looked at that didn't try to slate one of the other parties.

  About the only downside is that they don't eat their own dog food: after bringing through an Open Source Bill in the government of the Territory (another one), their web page appears to expect us to use a Microsoft non-standard character set. Not surprisingly, their web page doesn't validate, not even after guessing the character set. It's also sad to note that the Democrats don't mention IT policy on their policies page.

• http://www.greens.org.au/ have more in common with the Liberals than I think they'd like to admit: they have a policies page with lots of PDFs, but no overview.

It looks as if we have a long way to go before Australian politics really enters the Information Age.

Feedback from the Australian Labor Party today: the web site http://www.labor.org.au/ is not the official web site of the Labor party. For that, you need to look at http://www.alp.org.au/, not an immediately obvious name. So I did that and did find a number of policies, nicely marked “Election Policy”. Makes you wonder what the difference between an Election Policy and a non-Election Policy is. Unfortunately, it's still too detailed: I still can't find an overall statement of Labor's policy and values. I asked the web master why they didn't have a link at http://www.labor.org.au/, and he replied that they were separate organizations (he did, though, admit that it wasn't much of a reason).

Fewer concurrent things to do today; the big one was to get ARLA up and running on NetBSD, which required an all-day build to get to NetBSD-CURRENT. Took advantage of the time to finally migrate wantadilla, a task I had begun over two months ago. I still don't have my instructions for system upgrade complete, but the time was ripe, so swapped things around.

It would be nice to say that things went smoothly, but of course they did not. I had a lot of trouble starting X. First, I had forgotten that USB was disabled, so I couldn't get the mouse to work until I realized what the problem was, shut down, reconfigured the BIOS and then rebooted. Then I still had a lot of trouble getting the X displays to work. Given my X configuration, that's a little advanced, but in the end had to give up the attempt to move to new display cards (SiS) and revert to the cards I was using in old wantadilla. All in all, I was down for about three hours. The good news is that almost none of this had to do with upgrading the software: it was new hardware that (without good reason) caused me the headaches.

A whole day filled with meetings! Left at 7:30 am for Adelaide, where we had a meeting of the Board of Management of the IT Council of South Australia. For me the big issue was the membership committee, which met after the board meeting (not immediately after, as I should have liked; we ran short of time as a result). Discussed lots of things, but didn't get as far as we would have liked.

Lunch with David Newall in the “Corner Bistrot” [now apparently closed] in Bank St. Nice enough food, but I got the (incorrect) impression that they're pretty new, and they need more practice. Spent the time discussing the upcoming UNIX Developer's Symposium that we're trying to put together. It looks like it could be a real success.

Then back to the Radisson Playford for a meeting of the full IT Council; I get the impression that the effort was showing, and we finished early. With the current membership figures, the meeting looked a little superfluous. All that will change when we have our projected sharp increase in membership, of course.

Didn't get home until 18:30, to discover that my NetBSD build had crashed, probably as a result of the nightly crash of zaphod. High time to sort that out.

Yet another occasion where I find that simply building an updated version of the system takes forever. chili, the NetBSD -CURRENT system I started earlier in the week, has been running a 2.0H kernel since Wednesday, but building userland keeps causing problems. Yesterday it was because of the daily zaphod crash, and today NFS kept hanging. Finally decided to build it on the local disk (filling up a partition in the process), and experienced further hangs. I'd really expect Ethernet networking to be more reliable.

On the other hand, moving to the new wantadilla was less harrowing than I had expected. It seems that postfix's pseudo-sendmail program has changed its syntax and now requires the names of the recipients on the invocation line (previously it was happy to find them in the headers). Possibly there's a command line option to do it, but for now I just added the names to the command line:

Apart from that, some ports to add, and some configuration files I haven't saved yet. I've kept the old wantadilla running as blackwater (a word play), which makes that aspect easy, but is otherwise a mixed blessing: a number of my systems still have the old file systems on blackwater mounted, not the new ones on wantadilla. The background here is that I also migrated the IP addresses, and blackwater has kept the old addresses. As a result, the wrong files were getting updated from the kitchen and the brewing fridge .

I've got used to zaphod panicking every night, but this morning was different: after rebooting it wouldn't connect to the network. Everything else seemed OK. I changed the network cards, cable and even the switch to no avail, but I just got timeouts on the network connection (which was otherwise showing normal values). In the end moved the disks to the CVR machine that I had started working on where everything worked fine. Looks like the motherboard is on the way out. Now the question: do the panics have anything to do with that?

Spent most of the day working on AUUGN, but somehow didn't get very much done. One of the articles is a trimmed version of this diary, and I spent an inordinate amount of time getting the format right. I need to find a better way.

As expected, the new zaphod machine crashed as before this morning. I really must get round to finding out what's wrong here.

Spent most of the day working on AUUGN. Last time round seemed so easy, but this time it's keeping me busy. Maybe it's because I was no longer expecting it to be much work.

There's been this funny smell of overheating boards in the Mike Smith Memorial Room for a couple of days now. Nothing to be really worried about, but I spent some time wandering around sniffing, and found nothing.

This morning, after the obligatory reboot of zaphod, NFS didn't come unstuck. Further examination showed that blackwater, the old wantadilla, had had a power supply fan failure and (apparently) shut down due to overheating. Spent more time rebuilding, and then sent Yana off to town to buy more hardware. I've had a really bad time of it lately.

Spent the rest of the day working on the zaphod panic, which proves to be much more complicated than I had expected. Also, finally, chili (the new NetBSD test box) was upgraded to 2.0-CURRENT, and spent some time trying to compile ARLA, not helped by header file hell. How I hate GNU configure!

Still I can't get away from hardware problems. Today started with a relatively minor one: the CPU fan on the new wantadilla (all of 2½ months old!) started making noise. Confirmed that it was still operating correctly, so decided to do what most people do anyway and put the covers on the case: I normally have this tendency to leave the covers off, as the photos of the Mike Smith Memorial Room show. That silenced the noise.

One of the things that Yana brought back yesterday was a new (well, different) SMP motherboard that I had bought from somebody in town. Put that in the old zaphod case and discovered that it didn't power on. The lack of manual didn't help, and I spent over an hour before I decided that it wasn't going to work. Grrrr. The relationship between hardware price and the cost of time is changing to the point where I don't know why I bother.

On a more positive tack, managed to finally get ARLA to compile and “run” on NetBSD: I don't really know how to use it. Tried to build the documentation, but that seems to be broken too: it tries to reference an undefined file.

In the evening, just before bed, came in to find wantadilla dead. Further investigation showed it to have been overheating: when I rebooted it, the CPU temperature was still over 60°. Still further investigation linked that with this morning's well-intended replacement of the chassis covers. Took them off again, and the temperature dropped rapidly. The following screen shot was taken some minutes after it was back up and running ago. It shows the CPU temperature dropping from about 57° to under 50°:

Image title: temp Dimensions: 460 x 206, 6 kB Dimensions of original: 460 x 206, 6 kB Display this image: thumbnail    hidden   alone on page Display all images on this page as: thumbnails    this size Show for Tuesday, 28 September 2004: thumbnails    small images    diary entry

I wonder how much effort has really gone into investigating air flow in computer cabinets.

Spent the day chasing up ARLA; somebody in the FreeBSD project had assigned a problem report to me yesterday, and I thought it was a new version of the port. but it's one I've seen already, and it doesn't build. Didn't have the stomach to try to debug it, so read some of the documentation instead; that could certainly do with improvement. In any case, it doesn't look as if I can do much with only a single system running ARLA.

Also continued on the panicking zaphod problem; it looks like I bit off more than I could chew here, though if I get it finished it will certainly be a good one. It's now fairly clear that the bug is related to corruption on the disk. I had suspected this all along, but now I have evidence. It seems that I had put a “do not fsck” entry into /etc/fstab for this disk:

# Device                Mountpoint      FStype  Options         Dump    Pass#
# echunga:/src          /src            nfs     rw              0       0
/dev/ad1h               /src            ufs     rw              0       0

The commented out line is the clue: this was originally an NFS file system, which doesn't need fsck, and when changing it I forgot to set the fsck pass number (should be 2). As a result, I have now had the following corruption, probably since a month or so:

I could fix it, of course, and I'm sure that the problem would go away, but then I'd no longer have a case study.

Into Adelaide today for a couple of meetings. I had originally planned them around a renewed attempt to have an ADUUG lunch, but once again it had been cancelled. As a result to an interesting discussion with Jim Johnson in the morning, then had a bit of time on my hands which I spent with Tim Aslat. He had sold me a dual CPU motherboard on Monday, but I hadn't been able to get it to power on. Brought it and the old zaphod in with me, and of course it powered up with no problems doing exactly what I had done on Tuesday. I hate looking like a fool, but at least it works.

In the afternoon had a face-to-face status meeting instead of the usual teleconference. We've hired a whole lot of new people, and one of them, Peter Pham, is going to be working in the storage systems area for a while.

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.