|
| Greg's diary |
| September 2000 |
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Friday, 1 September 2000 | Echunga | |
| Top of page | ||
| next day | ||
| last day |
First day of Spring! And we got 22 mm rain. For once, I wasn't unhappy about that.
First thing this morning I went to find out why the house water pump (a Davey XJ70, sold to me only a little over a year ago as one of the most reliable pumps available) wasn't working. The results were depressing: the housing of the pump had cracked, and it was spraying water everywhere. That in itself wasn't bad, since the thing's still under guarantee, and I had was able to put the sprinkler pump in as a replacement, but it had drained our 70,000 litre water tank. I fear we're going to have problems with water this summer. Switched the toilets over to run on dam water.
Spent most of today trying to install IPSEC and raccoon, an IKE program. This was severely hampered by lack of documentation. I've decided I don't know what I'm doing, and I need to find some docco rather than waste time like this.
Also downloaded the latest patches for SMPng from Jason Evans. We really need to get this under CVS control. The official reason that we haven't is "it's not stable enough to go into the -CURRENT kernel yet, but it should be in a week or so, so it's not worth setting up a separate repo". The trouble is, it's been like that for a month, and we don't seem to be sticking to our agreement on what base the patches relate to. Ah well, with a bit of messing around I managed to get a kernel that built. And the &*&(*&(* wouldn't go into remote serial debug any more.
| Saturday, 2 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Weekend! Time for, well, work.
Yana had finally planned to update her sorely outdated home page today, but she had lost the magazine with the beginner's guide “How to write your own home page”. Took a look on Yahoo! and found a number of tutorials, but nothing which was as good a beginner's introduction. Oh well, she lost it, she'll have to put up with the consequences. I downloaded the Maricopa Center Writing HTML tutorial [as of February 2015, this link is defunct] and revised it a little. “Make sure you use consistent extensions”, they say. So the top level one was called start.htm and linked to tut/index.htm, which wasn't there... it was called tut/index.html. Still, it advocates using a text editor for writing HTML, but the pages they supply were all automatically generated with one of those emetic “single line per paragraph” editors. Wrote an Emacs macro to tidy things up a little, and while I was at it made a few changes to make it more suitable for UNIX.
Also played around a little half-heartedly with SMPng. Jason's been having trouble with serial debug too, could only get it to work at 9600 bps. While fetching a Coopers' it dawned on me: I disabled spls a couple of weeks ago, and didn't do any debugging since then. But I was the person who put the spl in getpacket in the first place, because the interrupt handlers were stealing characters meant for the debugger. Fixed that, I hope. It'll stop the clock, but I must remember to synchronize the clock after leaving the debugger.
| Sunday, 3 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another day with no real work (well, I've heard that's the way it's supposed to be). It became painfully obvious that Yana didn't understand too much about the UNIX file system, so I ended up writing a tutorial for her. Doubtless it can do with improvement.
Played around with the SMPng debug code a bit, not with much motivation. It seems to be more complicated than I thought: somehow the system never makes it as far as the debugger. Maybe it's somebody else's fault after all.
Finally got round to committing my first ever port, of x2x.
| Monday, 4 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Monday, bloody Monday! Some days start off bad and never improve. Today it started at 8:28 am, when I lost in an auction of an antique Boehm System Oboe. Grrr.
In the morning, back to trying to find out about IPSEC. There seems to be no docco anywhere worth having; looks like I'll have to shelve this project until I find some. On the way found some stuff about setting up IPv6, and I now have a tunnel from echunga. The following output is trimmed to show just the interesting stuff:
=== root@echunga (/dev/ttyp0) ~ 32 -> ifconfig -axl0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500 inet 192.109.197.82 netmask 0xffffff80 broadcast 192.109.197.127 inet6 fe80::250:daff:fecf:17d3%xl0 prefixlen 64 scopeid 0x1 ether 00:50:da:cf:17:d3 media: autoselect (100baseTX) status: active supported media: autoselect 100baseTX <full-duplex> 100baseTX 10baseT/UTP <full-duplex> 10baseT/UTP 100baseTX <hw-loopback> ed1: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500 inet 192.109.197.137 netmask 0xffffff80 broadcast 192.109.197.255 inet6 fe80::280:adff:feb7:c9c7%ed1 prefixlen 64 scopeid 0x3 ether 00:80:ad:b7:c9:c7 faith0: flags=8000<MULTICAST> mtu 1500 gif0: flags=8011<UP,POINTOPOINT,MULTICAST> mtu 1280 inet6 fe80::250:daff:fecf:17d3%gif0 —> :: prefixlen 64 scopeid 0x5 inet6 3ffe:b00:c18:1fff::4c3 —> 3ffe:b00:c18:1fff::4c2 prefixlen 127 inet6 3ffe:b00:c18:1fff::5bf —> 3ffe:b00:c18:1fff::5be prefixlen 127 lo0: flags=8049<UP,LOOPBACK,RUNNING,MULTICAST> mtu 16384 inet6 fe80::1%lo0 prefixlen 64 scopeid 0x9 inet6 ::1 prefixlen 128 inet 127.0.0.1 netmask 0xff000000=== root@echunga (/dev/ttyp0) ~ 33 -> netstat -rRouting tables Internet: Destination Gateway Flags Netif Expire default Cont0.way3.Adelaid UGSc 56 13079 ppp0 localhost localhost UH 10 1081 lo0 Cont0.way3.Adelaid echunga UH 54 0 ppp0 widecast/25 link#1 UC 0 0 xl0 => dialup48 link#1 UHLW 0 1 xl0 => zaphod link#1 UHLW 2 2051 xl0 => wantadilla 0:50:da:cf:17:97 UHLW 23 9039665 xl0 906 battunga 0:50:da:cf:7:35 UHLW 2 41264 xl0 1030 echunga 0:50:da:cf:17:d3 UHLW 10 30771 lo0 192.109.197.127 ff:ff:ff:ff:ff:ff UHLWb 1 410 xl0 192.109.197.128 ff:ff:ff:ff:ff:ff UHLWb 0 1 ed1 => widecast/25 link#3 UC 0 0 ed1 => echunga 0:80:ad:b7:c9:c7 UHLW 5 3728002 lo0 firefly.worldwide 0:60:97:40:fb:e1 UHLW 3 16717 ed1 466 broadcast ff:ff:ff:ff:ff:ff UHLWb 0 410 ed1 Internet6: Destination Gateway Flags Netif Expire default grog.au.freenet6.n UGSc gif0 ::1 ::1 UH lo0 3ffe:b00:c18:1fff: 3ffe:b00:c18:1fff: UH gif0 3ffe:b00:c18:1fff: ::1 UH lo0 3ffe:b00:c18:1fff: grog.au.freenet6.n UH gif0 grog.au.freenet6.n ::1 UH lo0 fe80::%xl0 link#1 UC xl0 fe80::%ed1 link#3 UC ed1 fe80::%gif0 link#5 UC gif0 fe80::250:daff:fec ::1 UH lo0 fe80::%lo0 fe80::1%lo0 Uc lo0 ff01:: ::1 U lo0 ff02::%xl0 link#1 UC xl0 ff02::%ed1 link#3 UC ed1 ff02::%gif0 link#5 UC gif0 ff02::%lo0 fe80::1%lo0 UC lo0=== root@echunga (/dev/ttyp0) ~ 34 -> ping6 www.6bone.netPING6(56=40+8+8 bytes) 3ffe:b00:c18:1fff::5bf —> 3ffe:b00:c18:1::10 16 bytes from 3ffe:b00:c18:1::10, icmp_seq=0 hlim=64 time=504.793 ms 16 bytes from 3ffe:b00:c18:1::10, icmp_seq=1 hlim=64 time=500.301 ms 16 bytes from 3ffe:b00:c18:1::10, icmp_seq=2 hlim=64 time=478.63 ms ^C —- www.6bone.net ping6 statistics —- 3 packets transmitted, 3 packets received, 0% packet loss round-trip min/avg/max = 478.63/494.574/504.793 ms You have new mail in /var/mail/grog=== root@echunga (/dev/ttyp0) ~ 35 -> traceroute6 www.6bone.nettraceroute6 to 6bone.net (3ffe:b00:c18:1::10) from 3ffe:b00:c18:1fff::5bf, 30 hops max, 12 byte packets 1 www.6bone.net 490.301 ms 497.482 ms 493.326 ms=== root@echunga (/dev/ttyp0) ~ 36 ->
Unfortunately I haven't been able to find any systems which aren't directly connected to the tunnel. Possibly I'm not understanding something there. Again, more docco would be of interest. For example, how do you register real IPv6 addresses?
Spent the rest of the day looking at the problems with remote serial debug. As far as I can tell, when I put the kernel into debug on boot, it says nothing, not even with remote serial debug speed set to 2400 bps. After a lot of frustration and too many kernel builds, I asked Jason Evans for one of his kernels, which showed exactly the same problem. Hardware? Well, no, a normal -CURRENT kernel works fine. Finally it transpired that Jason can't get it to go into the debugger on boot either.
Downloaded Yet Another Patch from Jason, and started building the world. While I was waiting, I found that Peter Wemm at Yahoo! has it up and running in multiuser mode on multiprocessors, and he's successfully building worlds (our “it's tough enough” criterion). Looks like we could commit the i386 code, but we're still behind with the alpha.
| Tuesday, 5 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
This SMP stuff is getting confusing. Spent a lot of time trying to work out why things aren't working here the way they are in the USA. I was able to establish that we can't handle serial debugging until the APIC is set up. If I attempt to go into debug at boot, the system just hangs. If I attempt to enter via the magic keyboard sequence ctrl-alt-esc, nothing happens until the APIC setup is complete.
In any case, this means that I still don't know if my fixes from Friday are correct or not. I'm reasonably sure they are, but for some reason my kernel won't go into multiuser mode, preferring to hang some time after starting init. I got hold of a kernel from Jason Evans, purportedly built from the same sources, and it ran fine. I'm suspecting some config file differences, and we'll have to look at that.
Here's the result of the buildworld:
chmod 444 freebsd.cf last command of the buildworld
6955.45 real 5369.47 user 2697.51 sys
18728 maximum resident set size
829 average shared memory size
848 average unshared data size
124 average unshared stack size
14923607 page reclaims
2462 page faults
0 swaps
708 block input operations
472310 block output operations
4137662 messages sent
4087072 messages received
8 signals received
5687003 voluntary context switches
6577958 involuntary context switches
This was working across NFS with -j4, and I can't recall how good that time is, but it's not obviously different from the old SMP kernel. Considering that buildworld is CPU-bound, that's about what you'd expect. Note that user and sys time together accounted for 8067 seconds in an elapsed 6955 seconds, or about 1.16 processors average. I'd have hoped for more, but this might be reasonable considering the NFS delays. What I do like is that we now have more detailed information about the CPU time spent in interrupt handlers:
=== root@zaphod (/dev/ttyp1) /src/FreeBSD/SMPng/src 5 -> ps aux
USER PID %CPU %MEM VSZ RSS TT STAT STARTED TIME COMMAND
root 11 48.3 0.0 0 0 ?? WL 6:12PM 774:58.07 (idle: cpu0)
root 10 48.0 0.0 0 0 ?? RL 6:12PM 775:03.34 (idle: cpu1)
root 1 0.0 0.2 532 212 ?? ILs 6:12PM 0:00.05 /sbin/init -s
root 12 0.0 0.0 0 0 ?? WL 6:12PM 6:59.36 (softinterrupt)
root 13 0.0 0.0 0 0 ?? WL 6:12PM 0:00.01 (irq14: ata0)
root 14 0.0 0.0 0 0 ?? WL 6:12PM 0:00.00 (irq15: ata1)
root 15 0.0 0.0 0 0 ?? WL 6:12PM 0:01.30 (irq10: ahc0)
root 16 0.0 0.0 0 0 ?? WL 6:12PM 0:00.00 (irq11: atapci1+)
root 17 0.0 0.0 0 0 ?? WL 6:12PM 0:00.00 (irq1: atkbd0)
root 18 0.0 0.0 0 0 ?? WL 6:12PM 0:00.00 (irq12: psm0)
root 19 0.0 0.0 0 0 ?? WL 6:12PM 0:00.00 (irq7: ppc0)
root 20 0.0 0.0 0 0 ?? WL 6:12PM 4:08.95 (irq0: clk)
root 2 0.0 0.0 0 0 ?? DL 6:12PM 0:00.44 (pagedaemon)
root 3 0.0 0.0 0 0 ?? DL 6:12PM 0:00.00 (vmdaemon)
root 4 0.0 0.0 0 0 ?? DL 6:12PM 0:00.33 (bufdaemon)
root 5 0.0 0.0 0 0 ?? DL 6:12PM 0:31.02 (syncer)
root 64 0.0 0.0 0 0 ?? WL 6:12PM 8:32.18 (irq3: dc0)
dc0, on irq 3, is the Ethernet driver, which had most of the work to do on an NFS build; by comparison, the disk drivers ata0 and ata1 did almost nothing, and the SCSI driver ahc0 had little to do. It's interesting to see how much time the clock interrupt (clk) and the soft interrupt handler have used. I've also been keeping an eye on the idle processes: last night it was looking as if the scheduler was favouring CPU 1, but this (Wednesday) morning's results look slightly the other way round. I suppose they're pretty much the same, anyway.
| Wednesday, 6 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
They've finally made a call for nominations for the FreeBSD Core Team Elections. I've nominated myself, because that's what they expect. I wonder if I should nominate somebody else. I also wonder who else will stand for election. I'm expecting a big shakeup in the core composition. The current team has 15 members, the new one will have 9. Of the current members, I don't expect to see more than about 4 in the new core team, and the only one I'm sure about is Jordan Hubbard. Personally, I will probably vote for four of the members.
Finally got my SMP kernels to run again. As I had suspected, it was a kernel configuration file issue: we don't support SMP on the 80386 processor, but I had both SMP and I386_CPU defined in the config file. Now we have an #error in machine/smp.h which catches that one.
Also got the debugger working properly again, modulo the problem getting it started on boot. As I suspected all along, it was the interrupt lockout issue.
| Thursday, 7 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
SMPng is committed! Now we can have lots of people pounding on it, and things should go quickly.
Started work on the SMPng paper for BSDCon. Considering that it's due by Sunday, it's about time. This is more difficult than I expected.
Peter Wemm claims that it takes about 30% longer to build a world on a quad Xeon with SMPng than it did before. That doesn't tie up with my measurements earlier, but my tests had been done over NFS. Tried to copy a source tree to the local machine, but I wasn't able to complete: the system hung up. Also, we still seem to have problems with serial debug, though it's a lot better.
Received confirmation from Malaysia: I'll be talking about BSD at the LinuxWorld conference in Kuala Lumpur in late November.
| Friday, 8 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
77 mm of rain in the last week! That's about 8% of the yearly rainfall, and just what the doctor ordered. I took at look at the water level in the tank today, and was more than a little pleased that it's about two-thirds full.
Surprisingly little feedback about the SMP code. Some people are reporting problems with serial I/O, which is interesting, because I don't think we changed much there. It's always been a “fast interrupt”, i.e. one which gets handled before reenabling interrupts.
Spent most of the day half-heartedly trying to find the motivation to write the paper on SMPng for presentation at BSDCon next month, but got dragged away to help with the serial port problems. It was pretty trivial: they hadn't been scheduling the soft interrupts.
Got brucified for my fix to sio.c. It seems that there's a subtle difference between schedsofttty and setsofttty which has far-reaching performance implications, but it wasn't documented anywhere. I will never understand why people are so opposed to commenting their code that they don't even want me to do it.
| Saturday, 9 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Oh, how difficult it is to get motivated for things you don't want to do. I really need to get this SMPng paper finished, but it's like pulling teeth. Did a few other things as well, but mainly procrastinated.
| Sunday, 10 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
I'm really leaving this paper till the last minute. It's got to be in by tomorrow, so I suppose I'll put in something and leave it as a URL for a work in progress.
Got a mail message from Leigh Hart suggesting I come and pick up a Sun 3/60 from him in exchange for a bottle of port. Since I had no time, went into town and picked up that, along with two monitors. Also picked up another 3/60, this time with only one monitor, from Justin Hawkins, who got an Exabyte 8200 for it. Leigh reminded me that he likes Viking Gold, apparently from the Barossa. I must write the name down, or I'm bound to forget and get him the wrong thing.
Leigh also showed me an interesting thing with his ISDN line. If you get Telstra's OnRamp Home Highway ISDN service, local voice calls cost $0.18 with no time limit. Data calls cost $1.00 per hour, which makes them still too expensive for permanent connection. This is stupid, monopolistic pricing, of course: data calls don't cost Telstra any more than voice calls. The only difference is the encoding performed at the end and the information in the call setup packets. So Leigh is setting the voice bit (DOV, Data over Voice), and presto! he's getting unmetered calls. You need an ISP who will go along with this, of course, and I don't suppose that would be Telstra. Interesting that Leigh works for Internode, who of course offer the service.
| Monday, 11 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Finally finished the SMPng paper. I'm not very happy with it. Probably the best part is the end of the abstract:
This paper is a snapshot of a work in progress. An up-to-date version is available at http://www.lemis.com/SMPng/.
I've given up on getting IPSec working on FreeBSD, at least until I understand what I'm doing. Started installing Debian Linux 2.2 on one of my boxes, so that I can cajole Martin into helping me install FreeS/WAN.
Hey, the Debian install worked fine this time, modulo some minor problems installing the Ethernet card. I really don't understand why every little piece needs installation separately; the docco says that the kernel is larger than normal because it contains support for just about anything, but it didn't recognize my Macronix Ethernet card until I found the correct module to load for it.
| Tuesday, 12 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Now that the SMP stuff has hit -CURRENT, activity is hotting up. Mark Murray already added the first new mutex, for malloc. I'm concerned that we're going to end up with a real mess of different mutexes for different things, and the deadlocks will become a real problem. At the moment it's keeping me busy on mail, anyway.
Completed the Debian installation, sort of. It's still confusing how you have to install every little thing. I seem to recall that earlier versions of Debian did the opposite and installed everything, but I'm not sure this is any easier.
Called up Martin Schwenke and discussed installing FreeS/WAN. He walked me through the Debian kernel build, which took quite a while because we had to work out which packages to install. It seems that Debian needs better grouping of packages.
Finally got all the stuff in place. FreeS/WAN expects the kernel sources to be in /usr/src/linux, though the recommendation I have had is that they should be in a subdirectory such as /usr/src/linux/kernel-2.2.17/. Anyway, mine are NFS mounted at /src/Linux/deviant/kernel-source-2.2.17/, so I was able to do a symlink. It ran for a while but died anyway. Investigation tomorrow.
| Wednesday, 13 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
This SMP stuff is still keeping me busy. It's surprising how many primitives there are, and how much people's understanding of them is coloured by their background. For example, we weren't able to establish whether there's a difference between Dijkstra's semaphores and counting semaphores. At least one description of the “Dijkstra” semaphores suggests that they don't have a counter, but others do, and there's circumstantial evidence, like the fact that the Algol 68 sema type includes a counter. From a purely implementation point of view, you need some variable to specify whether a semaphore is available or not, so it may as well be a counter. In addition, Dijkstra uses different names for getting and releasing a semaphore (P and V, which I learn today come from the Dutch passeren and vrijgeven), while others call the same functions wait and post. Sigh.
Spent a lot of time today writing a document summarizing the differences, but I'm still not happy enough with it to send it out to anybody. SMP-related mail is still taking up a surprising amount of time.
Found my problem with the FreeS/WAN installation. For some reason it did a kernel build in a different manner, and it tried to invoke a program as86 which wasn't installed. OK, I'm getting used to dselect now, so after only a few minutes I decided it had to be in the nasm package. So I chose to install it. dselect found some other goodies to go with it:
The following NEW packages will be installed: autoconf automake biff bin86 bison ccmalloc chdrv chdrvfont cjk-latex cvs cvs-buildpackage ddd debconf dialog doc-debian doc-linux-text dpkg-ftp electric-fence finger fingerd flex freetype-tools freewnn-common freewnn-cserver fvwm-common gcc-doc gettext glibc-doc gnotepad+-help gpm iamerican ibritish indent lclint lesstif1 libdigest-md5-perl libltdl0 libltdl0-dev libnet-telnet-perl libtimedate-perl libtool lpr lsof-2.2 m4 mailtools manpages-dev memprof metamail mime-support mpack mtools mutt nasm nfs-common nfs-kernel-server perl-5.005-doc perl-5.005-suid pidentd postgresql-slink procmail python-gtk rcs sharutils stl-manual strace talk talkd task-debian-devel tclreadline tcsh tcsh-i18n telnetd texinfo tfm-arphic-bkai00mp time transfig ttf-arphic-gbsn00lp wenglish xa+cv xcin2.3 xcingb xfonts-arphic-gbsn00lp
Wow, that's a lot of dependencies. Why do I need mutt or fvwm in order to run the assembler? And when installing CVS it took a lot of convincing not to set up another repo on this machine. Anyway, I was right, as86 was in one of the packages, so off to run make. It seems to start from scratch.
After that I was able to build and install a new kernel, and FreeS/WAN came up, if that's the right word. Something in the installation also killed my mouse. Something to investigate tomorrow. Spent some time on the phone with Martin Schwenke discussing the config files, and nearly got finished before Martin had to leave. This 1½ hour time difference is more of a nuisance than I expected.
| Thursday, 14 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Some days just nothing works. I spent half the day catching up on mail, and the other half trying to get FreeS/WAN up and running. Yesterday evening it was almost working. Progress: this evening it was almost working. I'm reminded painfully that I still don't understand how this stuff works.
Mail is getting to be a real problem. I still haven't thought the SMP locking primitives stuff through, but I managed to receive 1084 mail messages. Fortunately I can delete most of them unread.
| Friday, 15 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Somebody has got hold of my email address and sold it to the spammers. OK, no mercy. I wrote a script to:
All I need to do with spam now is to pipe it to blacklist.
This script is now lost in the mists of time (10 September 2008)
Spent the afternoon again trying to get FreeS/WAN up and running. It isn't as if we didn't make any progress, more like we weren't sure in which direction. We got it as far as being able to ping, but only after timeouts whacking. Martin is still 1½ hours ahead of me, so he had to leave, and I've decided to shelve it—again—and try a local setup first. Maybe I'll even go the FreeBSD IPSec way in the process.
| Saturday, 16 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
For a change I remembered that it's a weekend and didn't do too much today. Spent a bit of time building a release CD-ROM for SMPng on my laptop, so that I can demonstrate it on my coming travels. We don't really have good build targets for just building a system and installing it on a second partition, so it look as if I'll have to burn a CD-R.
Also did some thinking about presentations.
Apart from that, also did some thinking about playing in the Messiah this Christmas. It's straightforward enough music, so I thought Yana and I could both sign up, but on examining the score I found that it doesn't include flutes. Dragged out Yana's neglected oboe and found that we still have a couple of functional reeds. Maybe we can make something of that.
| Sunday, 17 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
No work! Holidays are coming, and I had other things to do. Actually got a bit of music practice done.
| Monday, 18 September 2000 | Echunga –> Hawker | Images for 18 September 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up early this morning to find out if we could leave for Lake Eyre today rather than tomorrow. The main issue was the availability of the Landcruiser, which we had booked for tomorrow. We had planned to leave tomorrow after the insurance valuer came, which would mean leaving Adelaide at about 1 pm. Yvonne hadn't checked the distance to William Creek: it's about 900 km from Adelaide, about 250 km of which are on dirt roads, so it would be almost impossible to get there.
Things worked out: we had to take Yana to the masseur at 11:15, and they said they'd have the car ready by 1 pm, so we set off there, meeting Richard Sharpe at the masseur. Things worked out relatively well: the car was ready when we arrived, and we left without any serious problems.
Off North to the Flinders Ranges:
We arrived in Quorn round 5 pm. We had booked a hotel in Hawker, but instead of going there immediately, detoured off the road to see Warren Gorge, which was worth while.
Arrived in Hawker round 7 pm, and checked into the pub. Had a reasonable dinner at the “Ghan”, a converted railway station. About the only other guests were some Germans with powerful southern accents. We couldn't quite make up our minds whether they were Swabians or Bavarians. The place seems to be full of Germans.
|
|||||||||||||||||||||||
| Tuesday, 19 September 2000 | Hawker –> William Creek | Images for 19 September 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up bright and early this morning and off on a slight detour to see Wilpena Pound. Stopped on the way to take some photos and for Yana to look for a necklace she had left in Hawker:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
I don't know what this flower is, but there were quite a few of them around. I'd be grateful for pointers.
|
|||||||||||||||||||||||
I do know this one now. Pretty, isn't it? It's Echium plantagineum, an introduced pest called Salvation Jane (called “Patterson's Curse”, or even “Patterson?s Curse” on the East Coast), and it's the scourge of farmers all over the state. It can completely take over a paddock, so that nothing else grows, and though stock can eat it, it doesn't do them any good over long periods of time.
At Wilpena Pound we found too much tourism for our liking, along with the information that no horse riding was allowed within the pound, so we didn't stay there long. It looks better from outside:
|
|||||||||||||||||||||||
After that, on through Brachina Gorge, which is interesting because of the ancient geological structures, going from 640 million years old at the entrance to 520 million years old at the other end. Round about this time, we discovered that the air conditioner had failed on us.
|
|||||||||||||||||||||||
From there, on through Leigh Creek, Lyndhurst to Maree, a distance of about 200 km. These weren't just the most important towns on the road, they were the only towns on the road.
|
|||||||||||||||||||||||
|
|
|||||||||
|
|||||||||||||||||||||||
Maree is the junction of the Birdsville track, which leads up into Queensland, and the Oodnadatta track, which follows the old route of the transcontinental railway (the “Ghan”, after the supposed Afghans who even earlier transported by camel) up towards Darwin. That's the one we took.
Yvonne was a little concerned about the appearance of Maree, which can't have more than 1000 inhabitants,
In fact it's about 150.
and was anxious to get on to William Creek, 202 km away. Well, there wasn't much in our way—William Creek is the next town—but it was pretty heavy going. On the way we drove past South Lake Eyre, but there wasn't much water to be seen.
About the first interesting thing was just before William Creek:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Feral cats are a real pest in the Outback. At the general store in William Creek they had an exhibit showing how many native animals a single cat can eat. I'm a cat lover myself, but it's clear that feral cats and domestic cats are two completely different propositions.
Yvonne was delighted when we got to William Creek. It's the prototypical bush town in the middle of nowhere, population 16.
| Wednesday, 20 September 2000 | William Creek –> Coober Pedy | Images for 20 September 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up bright and early this morning to get to Lake Eyre before it got too hot, especially in view of the fact that the air conditioning had packed in. We went down to the William Creek Hotel for breakfast at 7:30. By the time the food came, it was 8:10, and the bacon and eggs (if that's what they were supposed to be) were cold, greasy and burnt. I sent them back before taking a photo, but that didn't make much difference, because the replacement looked only fractionally better:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
I was seriously considering refusing to pay for the food, but I didn't need to. The owners of the pub knew of the problem, apologized and said we didn't have to pay. “He can't cook eggs to save his life”. It's a good thing, too. They would have charged us $9.75 each for the breakfast. I suspect they didn't know that there were four of us.
Off to Lake Eyre round 9:15, and bumped into a ranger at the do-it-yourself desert pass station, and got some tips from him, including the information that the road wasn't too bad, and that plenty of people did it in two-wheel drive vehicles.
Half way to the lake, we found the place where an Austrian tourist had died of thirst on 13 December 1998:
|
|||||||||||||||||||||||
When we got to the lake, we found two different places to look at. The first one was closer, but we spent half an hour walking out on the salt flats looking for water. After a while, the salt got too soft, so we turned back. The following photo of Yvonne sets the mood: the flies were really unbelievably bad.
|
|||||||||||||||||||||||
Next we set off to Halligan's bay. For the first time, the surface was bad enough to need four-wheel drive: sand. I put the car into four-wheel drive and got bogged down anyway, in my white Toyota Landcruiser. This is very possibly exactly the place where Göschka and Großmüller got bogged down.
We were luckier. I hopped out of the car and flagged down the next car coming the other way, about 10 seconds later. He hopped out, took a look at the car, turned something on the hubs of the front wheels, and drove it out. By the time he was done, about 6 cars had gathered to watch the fun. It seems that the Landcruisers, and possibly most other truck-like four-wheel-drive vehicles, have a separate mechanism for disengaging the front drive shafts when not in four-wheel-drive mode. Without that, the car was just a two-wheel-drive truck, and more likely to get bogged down than a lighter vehicle. Did the bloke who hired the car to me tell me about this? Not a word.
That begs the question, doesn't it? Göschka and Großmüller were in pretty much the same car, and they got bogged down in sand which a Landcruiser should take in its stride—if they have the drive shafts engaged. I wonder if the rental people don't have a serious part to play in the death.
Halligan's Bay itself was disappointing. No birds, apart from a surprising number of seagulls, no flowers, nothing much except for a sign saying that we had reached rock bottom: this is the lowest place in Australia, 16 metres below sea level.
On the way back, we noticed dozens of tennis balls lying on the side of the road. On closer examination, they appear to be some kind of fruit, melons or pumpkins. I have no idea what they really are.
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
When we got back to William Creek, we had had enough. We were covered in dust (thanks, air conditioning), and there wasn't much more to do. After the disappointment of seeing the lake from the surface, we decided against staying another night to spend $520 for a one-hour scenic flight over the lake, and tried to check out and head off to Coober Pedy. To our surprise, though it was 2 pm, we had no problems, so set off on our last 160 km of dirt roads to Coober Pedy and relative civilization through the Woomera Prohibited area.
|
|||||||||||||||||||||||
In the middle of this area, we came across what might be the world's longest barrier, far exceeding the Great Wall of China:
|
|||||||||||||||||||||||
|
|
|||||||||
In Coober Pedy, we went to the hotel we had booked. That's the last time I rely on Lonely Planet: the place was terrible. We found an excuse and went across the road to another place which was both cheaper and better. It wasn't quite underground (in fact, I don't think there is much stuff underground), but it was piled up in “underground” style. Underneath was a real underground backpacker's hostel, which I got on video, but not on film.
Coober Pedy's a funny place. The first hotel was run by a woman whom I suspect to be Polish or Russian; the second one was run by a bloke who could be Turkish. We had Greek food in the evening.
| Thursday, 21 September 2000 | Coober Pedy –> Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
It wasn't difficult to get up early this morning. I don't know what it is about backpackers, but some came in in the middle of the night and had a loud conversation while running up and down the stairs next to our room. After they had run out of steam, a family including a small boy started shouting (maybe they would call it talking) directly outside our bedroom. I finally staggered to my feet, thinking it was morning, and found it was 4:55 am. I said “It's five o'clock in the bloody morning. Can't you be quiet?”. The little boy was impressed: in the same tone he called “Mum, that man says it's 5 o'clock in the morning”. At least his mother had the good sense to quieten him down.
Had breakfast at the “Last Resort Cafe”, run by a Swiss woman, with whom Yvonne spoke in German. The bacon and eggs were good, but by this time our expectations weren't too high anyway. I found this gem in a brochure I had picked up on Central Australia:
You could call Alice Springs and region the gastronomic capital of Australia.
Well, yes. I suppose you could call black white as well. They appeared to be deathly serious.
After breakfast, we went round an old opal mine in town. Mining has been prohibited inside the city limits of Coober Pedy, so they turned the place into a gold mine (self-guided tourist visits) instead. The most interesting thing about it was that while building the tourist trap, they found lots of opal in the walls, some of which was put on display.
Then off on the long trek home. About 900 km, but all on sealed roads. Drove into Woomera (just off the highway) for lunch, and discovered that none of my passengers had ever heard of it.
Coming into Port Augusta we could see the Flinders Ranges in mid-distance off to the left. On return, I checked and found that they were about 75 km away. It's amazing how clear things look in the desert.
It's really interesting how different things suddenly became at Port Augusta. After hundreds of kilometres of nothing, we were suddenly back in a familiar environment. The pollution caught up with us, too: going down to Adelaide, the North Mount Lofty Ranges were only 50 km away, but they were much hazier.
Back home without any further ado, arriving at about 8:15 pm.
| Friday, 22 September 2000 | Echunga | Images for 22 September 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Ah, it's good to be home. Finding 5,000 mail messages waiting for me didn't help much, though, and I spend most of the day reading mail or mowing the lawn. Spring is in the air, it's stopped raining, and it's just too much for me to be able to stay inside.
| Saturday, 23 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent much of the day writing up this diary. I'm becoming intrigued about the death of Caroline Großmüller. I talked to a 4WD freak friend of mine in Utah, and he told me:
I see the concept of locking hubs isn't part of the standard lore in your neck of the woods. Every 5-year-old boy here knows what locking hubs are, and which way to turn them, even if they're not physically capable of doing so. Interestingly enough, Toyota hasn't sold LandCruisers with locking hubs here for a decade, they're all full-time 4wd.
Assuming that they really only got bogged down because the car wasn't in 4WD mode, this suggests that both the manufacturer and the renter have been guilty of negligence in this matter.
I really need to get my docco up to scratch before the month.
| Sunday, 24 September 2000 | Echunga | Images for 24 September 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
It's interesting keeping a diary like this. It shows how irregularly I work: sometimes I do lots of work in a short period of time, sometimes I do very little and take a long time to do it. OK, today was Sunday, so I should have been taking it easy.
Tried yet again to build a release of SMPng. Somehow something always goes wrong. Today, after 4 hours, it decided that it couldn't find install(8), which is just plain nonsense. Decided it might be an NFS problem—we still have them—and checked out locally (onto a conveniently large Vinum volume). Another day's work for zaphod.
Spring is coming, and it looks like we won't have to heat for a while. Saddled up Darah and Shaleema and set off for Kuitpo Forest. We hadn't got more than 500 metres down Duffield Road before Shaleema lost a shoe—again—and we had to turn back. That makes three times in the last four rides that she's had trouble with the same shoe. I wonder if we should investigate another farrier. Went back into the arena and worked Darah for a while. She needed it: it must be over a month since I last rode her, and the weather's only partly to blame.
Started (finally!) working on next month's Dæmon News article. History of Core Teams.
| Monday, 25 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Martin Nightingale is leaving us, to my personal sadness. He'll be in Canberra to say goodbye on Thursday, and so will Angus, so I suppose I should go there too. In view of the short notice, decided to do a flight booking via the web.
Two hours later, and after three detailed messages to the airline's webmaster (none of which were replied to on the same day), I gave up. The software is just plain broken. What a waste of time. I couldn't even update my frequent flyer profile correctly: the field for entering the phone number is too short. It looks like it'll be a long time before the Web really becomes mainstream.
More docco. It's funny how these things start: it's like pushing a stalled truck over the crest of a hill. At the beginning, you really have to push to get anything done, and after a while it takes off by itself. I'm not quite that far yet, but the looming end of the month doesn't do any harm.
Another attempt to build a release on zaphod.lemis.com fails. This time it's apparently a system stability problem: the system apparently spontaneously reset. Grrr.
| Tuesday, 26 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Finally got some progress on the docco front, finishing both the Dæmon news article and the paper for Malaysia. It kept me going all day, of course, but it means that I can go to Canberra tomorrow without worrying about deadlines.
Building a release is still broken, in the documentation. At least Jordan has noticed now and reported it. Did a make -k and managed to complete it, but now I won't be able to burn a CD until I return from Canberra, so I might as well start again from scratch.
| Wednesday, 27 September 2000 | Echunga → Canberra → Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up bright and early this morning, and off to the airport for the 7:05 flight to Canberra. This 1½ hour time difference is somewhat painful.
I suppose I should come here more often. I don't get much work done, but it's good to stay in touch with the blokes in Canberra. It's a pity about the work I had planned to do today, but I suppose it's worth leaving the original plan below to remind me just how far apart plan and reality are.
One good thing was that I managed to palm off the FreeS/WAN stuff on Martin Schwenke. He didn't promise too much, but I can always pretend.
| Thursday, 28 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
I should travel more often. It gives me time to stop doing normal work and catch up with my mail.
A number of interesting discussions today, not too much of which I can put in this diary. One result was that I finally decided to come out of the closet and add myself to the web page at http://www.linuxcare.com.au/people/. Now to update my home page.
27 July 2004: Time has worn this diary entry particularly. For one thing, of course, the Linuxcare Australia web site is long gone. The other relates to the discussions “not too much of which I can put in this diary”. In fact, I believe this was the day I went to dinner with Tridge and Warren Toomey, and Tridge spent all evening talking about a new hack he was working on: getting a TiVo to work under Australian conditions. I suppose it's a sign of the times that I didn't even mention it.
| Friday, 29 September 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another day where I seem to get nothing done. Somehow I have to find a better solution to getting through my daily 1000 mail messages. Some times it seems to take all day and I get nothing else done.
I did get my make release finished today, but that didn't take much work. Apart from that, played around with my home page and brought it somewhat up to date. Now to look at the issues of creating presentations with groff.
| Saturday, 30 September 2000 | Echunga | |
| Top of page | ||
| previous day |
I suppose I might as well admit it that I can't always perform at full power. There are plenty of important things lying around, but sometimes I need to do the less important ones.
Finally got a useful make release result for SMPng, and burnt a CD-R. Only when I tried to boot from it did I realise that I didn't have a boot image. My first coaster, though strictly it's legible. Burnt another one and discovered that we still have problems with the kernel location, and I can only boot the image manually. Still, it was enough to install on sydney, my laptop. Unlike other people's experience, I had no trouble.
The following text was originally written on 26 September 2000, and the projection has now been taken over by reality. Still, it's interesting to compare planning and reality: Henry has asked for us to have our diaries up to date to the end of the month by Wednesday, so here goes...
Up bright and early this morning, and took another look at the Vinum root file system issue. It's much easier than I thought, and I had the code written, tested and committed before breakfast.
After breakfast, looked at the wish list. There was a fair amount of stuff there, and it kept me busy all day. We now have remotely accessible volumes via TCP.
After my success yesterday, decided it would be time to finally port Vinum to OpenBSD. In fact, it wasn't really that difficult, though it did take all day. The biggest thing was that OpenBSD still uses the old interfaces which we got rid of in FreeBSD version 3. Good thing I had #ifdefs in there for version 2. Things look a little wobbly, but it's running.
Well, if I can do it with OpenBSD, why not Linux? After all, I have a Linux machine right next to my old, mouldy test disks. I wish I understood Linux better, but by the evening I had most of the stuff in place. These Linux block device drivers are strange. Maybe tomorrow I'll port the FreeBSD CAM layer to Linux.
Do you have a comment about something I have written? This is a diary, not a “blog”, and there is deliberately no provision for directly adding comments. It's also not a vehicle for third-party content. But I welcome feedback and try to reply to all messages I receive. See the diary overview for more details. If you do send me a message relating to something I have written, please indicate whether you'd prefer me not to mention your name. Otherwise I'll assume that it's OK to do so.
| Top of page | Greg's home page | Today's diary entry | Next month | Greg's photos | Copyright information |
