Greg
Greg's diary
September 2003
Translate this page
Select day in September 2003:
Su Mo Tu We Th Fr Sa
1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30
Select month:
2003 May Jun Jul Aug
2003 Sep Oct Nov Dec
2004 Jan Feb Mar Apr
Today's diary entry
Diary index
About this diary
Previous month
Next month
Greg's home page
Greg's photos
Network link stats
Greg's other links
Copyright information
    
Groogle

Monday, 1 September 2003 Sydney
Top of page
next day
last day

Up bright and early this morning and next door for breakfast in a little place called the Fish Shop, where fish were conspicuously absent, and had a surprisingly expensive breakfast. Then to my kernel debugging tutorial, which was better visited than I expected. Despite all the work I did on the presentation, found numerous deficiencies. Ah well, the whole thing is all about debugging, and the tutorial evaluations were reasonably good.

In the evening, still no satellite downlink. Out looking for food, and decided that Milson's Point is not exactly a gourmet paradise. Had some reasonable Chinese/Thai food in a freezing little restaurant, and then discovered I was still hungry. Ended up eating an antipasto platter in the hotel bar, talking to Joel and Carl Makin, whom I have only just met after years of correspondence.


Tuesday, 2 September 2003 Sydney Images for 2 September 2003
Top of page
previous day
next day
last day

AUUG board meeting this morning, which I had tried to keep short, but somehow we managed to go on all morning. At any rate we managed to cover a fair amount of ground, in fact more than intended. It looks like we have a good board of directors this year, though tomorrow we'll have a different one from today: David Bullock is resigning, so we agreed to coopt Andrew Cowie to replace him.

In the afternoon spent most of the time writing up my President's Report for tomorrow's Annual General Meeting. Then in the evening to the speaker's reception, which was at an Indian restaurant with a nice view over the harbour (photos here). Pleasant evening.


Wednesday, 3 September 2003 Sydney
Top of page
previous day
next day
last day

And off to the start of a new conference! Got up pretty late, and didn't get breakfast until after opening the conference.

Later, Kieran O'Shaughnessy of SCO Australia came along for a panel session with Con Zymaris and myself. Not surprisingly, the session was well attended. Kieran brought out the party line with some ridiculous number of claimed lines of infringement in the Linux kernel, so I brought out a display of the BPF code of their second example presented on 18 August, then paged back to the beginning to show the BSD license there. The look on Kieran's face really made my day. He claimed, rather stupidly, that these were just examples of the techniques that they used to find similar code. Greg Rose joined in the fun and pointed out that some of his code, written in 1976 (still at least three years after the first example) had been included in UNIX without his consent, and asked whether Kieran thought he should approach IBM for royalties. Finally I got to do the summing up, where I determined that SCO had done nothing to show any abuse of its rights, but that the code in question had indicated that their code was way out of date, and that it looked likely that they had abused the BSD license (applause).

After that, a press interview which to which I'll link when it's published. Didn't do much in the afternoon. In the evening we had the networking reception, after which out to dinner with Steve Alford of the National Office for the Information Economy, along with David Purdue and Gordon Hubbard. Interesting story, amusingly a year to the day since our last dinner with Steve. Things have certainly changed a lot now, and we're getting to understand the terrain better.


Thursday, 4 September 2003 Sydney Images for 4 September 2003
Top of page
previous day
next day
last day

How tired I am! I wonder if it's just getting old, or whether the stress of the conference is too much.

Up bright and early to chair the keynote sessions from Steve Alford (NOIE) and Elizabeth Gordon-Werner from the NSW government. Both came across very well. Then off to fight a losing battle against my mail, now over 4000 messages and counting. In the afternoon there were a couple of interesting talks about kernel stuff on the Gelato project, then Phil Karn ranting about how Microsoft has broken the Internet. That was less interesting, but only because he said pretty much what I would have said.

Conference dinner in the evening, the usual madness. Also had the presentations of the Australian Open Source Awards, three of the four going to Ozlabs. This is getting boring, though we had a rather amusing presentation of the technology award: Martin Pool won it, but he wasn't there, so I got Martin Schwenke to accept it on his behalf. Put on a bit of a skit in the process (“Wait, I know you. You're Martin the Pool Champion, not Martin Pool the Champion”).

Then the room party, in my room, and despite my earlier tiredness, I managed to hold out. Photos are here. To bed just before 3 am.


Friday, 5 September 2003 Sydney Images for 5 September 2003
Top of page
previous day
next day
last day

Gordon woke me this morning at about 10:30 to tell me that there was a photographer downstairs to take a photo of me. I had been warned yesterday, but forgot about it, so downstairs in a hurry.

After that, had breakfast with David Newall, always a good choice for a late breakfast, then to pick up my mail, and discovered that echunga was down. Why do I always have this kind of problem during the conference? Last year I had problems with echunga wedging, and two years ago the UPS decided to refuse to supply power. Today was worse: we couldn't reboot echunga, and in the course of the afternoon we got Bernd Wulf to come over and try removing the SCSI host adaptor, which looked suspicious and wasn't needed anyway. After that, the machine booted, but couldn't get through the fscks without wedging again, so in the end put the disks in the NetBSD test box and left it like that. What a pain! Bernd noted that the electrolytic capacitors in the processor voltage regulators were bulging, a good sign of hardware problems, but one that was as good as unknown until a few years ago. Is this really just a problem of quality control? It's interesting to note that both echunga and wantadilla (which began failing less spectacularly a month or two ago) had Epox motherboards.

To make matters worse, the conference network got removed while Bernd was doing his thing, so I couldn't even check. This means I won't have network connectivity until Saturday in the USA.

Did my talk in the afternoon, and for once I thought I had done a reasonable job. The feedback forms seem to agree. That's rather gratifying when I think that I had deliberately taken an unpopular stance. Interestingly, it also showed a weakness of the software I'm using: ghostscript seems unable to maintain embedded image quality while converting PostScript to PDF, so went and talked to the Apple people, who did a good conversion after a lot of clicking and grunting.

Back to Monday's Indian restaurant in the evening, and somehow managed to make quite a late night of it. Thank God the stress is over.


Saturday, 6 September 2003 Sydney –> Los Angeles –> San Francisco –> Sunnyvale
Top of page
previous day
next day
last day

Woke up this morning late, feeling terrible. Off to the airport, where I seriously considered not flying, and finally to the medical centre to have myself looked at. They found nothing seriously wrong with me, so off onto the plane, where I gradually became more human. It must be the strain of the last few days. I suspect that the problems with echunga played a particularly significant role.

It's funny that I didn't go into more detail. My blood pressure was very high, systolic over 140, but nothing that looked like remaining that way.

Off to Los Angeles with no particular issues beyond a no-show having his bags taken off the plan. 30 minutes delay: I wonder how much that costs. Sure, he lost his fare, but I suspect that's only part of the cost.

Arrived in Los Angeles and to the President's Club, where I finally found somebody using the wireless networking. Based on the name, I thought he was accessing a porn site, but I've since been informed that it's a chat site for gay males looking for sexual partners. It seems that there's a big difference between the legal status of the two.

11:38:03.025267 web3.public.m4m4sex.com.http > 10.240.17.95.49965: S 4244509477:4244509477(0) ack 2611155728 win 5840 <mss 1460,nop,wscale 0> (DF)
11:38:03.148527 web3.public.m4m4sex.com.http > 10.240.17.95.49965: .  ack 665 win 6640 (DF)
11:38:03.190016 web3.public.m4m4sex.com.http > 10.240.17.95.49965: .  1:1461(1460) ack 665 win 6640 (DF)
11:38:03.197807 web3.public.m4m4sex.com.http > 10.240.17.95.49965: .  1461:2921(1460) ack 665 win 6640 (DF)
11:38:03.322397 web3.public.m4m4sex.com.http > 10.240.17.95.49965: FP 5841:6682(841) ack 665 win 6640 (DF)
11:38:03.331508 web3.public.m4m4sex.com.http > 10.240.17.95.49965: .  2921:4381(1460) ack 665 win 6640 (DF)
11:38:03.339396 web3.public.m4m4sex.com.http > 10.240.17.95.49965: .  4381:5841(1460) ack 665 win 6640 (DF)
11:38:03.473068 web3.public.m4m4sex.com.http > 10.240.17.95.49965: .  ack 666 win 6640 (DF)

Up to San Francisco, where the baggage handling people impressed me again with their inability. On the positive side, they've finished the rail connection to the rental car facility, which improves things. Hertz impressed me with a form which I was supposed to sign without reading, and which reprinted every 30 seconds just to be on time. What unbelievable behaviour!

Down to Silicon Valley and to Fry's, where I had vowed never again to set foot, and found good reason to regret having gone there again. It seems that 802.11 access points (sorry, “WiFi Hotspots”) are no longer in favour, and now you need a combined device with an Ethernet switch and an ADSL port. Left without buying anything and off to see Roberto Tomasi. On his advice, down to CompUSA, who had a smaller collection of the same junk, and later to the Sneha Indian restaurant on the corner of Central and Lawrence Expressways. Somehow Indian food has been the flavour of the week (Monday, Tuesday, Friday and today), but I really like this particular place, and I wasn't disappointed today. The fact that most of the people there were Indian was also a good sign.


Sunday, 7 September 2003 Sunnyvale –> San Mateo
Top of page
previous day
next day
last day

Another bad night's sleep, and up early as a result. Roberto doesn't eat breakfast, so off to Denny's for that, being reminded in the process how much sweet food the Americans eat: a “meat lover's breakfast” included three enormous pancakes and maple syrup, something I would normally never eat.

To Fry's again looking for 802.11 access points (or should I be saying “WiFi Hotspots”? No, I don't think so). It's amazing how the prices differ. After some deliberation, discovered that the AirLink 802.11g/802.11b routers cost the same as the AirLink 802.11g access points, and that they offered all the same features, but also a cable/ADSL gateway and a four-port Ethernet switch. Bought two, and discovered that they were on special offer at the same price as the cheapest 802.11b-only access points ($70). All would have been wonderful if they hadn't refused my credit card, so I had to pay cash.

Again missing details: at the same time I bought a blood pressure monitor as the result of the previous day's scare.

Up a very congested highway 101 to the Marriott hotel outside the San Francisco airport, where I arrived at midday to discover that it was the wrong hotel. Off with directions to the Marriott San Mateo, which must be the best-hidden place in the area. Finally checked in to find that they had changed my reservation, and that I was only booked in for 3 nights. They were able to change that, so the next problem didn't happen until I tried to drive round the hotel to my room: my key card hadn't been coded correctly.

Off to the airport and dropped off the car, barely making it back to domestic terminal 1 for the hotel shuttle pickup. The idea of the overhead train is good, but the implementation still needs tweaking. From the train stop to the hotel pickup took six minutes of wandering through the central car park. The only signs assumed that I wanted to fly somewhere. Barely made it on time (on the hour) for the pickup, but there was none to be seen. Called the hotel, and was told to wait downstairs, and that the pickup would be there on the half hour. That seemed improbable, but the girl (Shauna? she spoke very quietly, despite my requests) insisted on no less than four occasions. I finally asked a policewoman downstairs, who insisted that she would not allow a hotel shuttle to stop down there, so back and spoke to somebody else (Craig), who spent a lot of time checking and confirmed that the shuttle would arrive on the upper deck before 2:30 pm.

It didn't. At 2:45 I called yet again (sixth time for something which shouldn't need a phone call at all) and they apologized and told me they'd pay for a cab. This must be some kind of record: all in all, three hours to check in and drop off my car.

In the evening to conference registration, where I discovered that the registrations this year were less than half those of last year. And we thought we had trouble at the AUUG conference. Had a drink and something to eat with Greg Shapiro, Greg Skafte and Shane Kosowan, and off to bed early.


Monday, 8 September 2003 San Mateo
Top of page
previous day
next day
last day

Spent some time today playing with the wireless routers, and discovered a number of interesting things. One was that there are significant bugs in the firmware, but another is that the system is Linux based. Unfortunately, the manufacturer doesn't seem to have a web site; more investigation is needed.

First day of the BSDCon, and to Kirk McKusick's code reading tutorial; to my amusement, he too had submitted a draft as tutorial notes. Got some interesting information, not the least ideas for presentations.

In the evening with a few Swiss people to Palo Alto to Gordon Biersch, a microbrewery. To my astonishment, they had a Weizenbier which tasted remarkably authentic. Much talk, but unfortunately little understanding due to the high noise level. Took a number of photos.


Tuesday, 9 September 2003 San Mateo Images for 9 September 2003
Top of page
previous day
next day
last day

Up for my own tutorial today (kernel debugging), relatively confident that things would work OK, since I did it only last week. It did, indeed, work OK, though it's astounding how differently it went from last week. In particular, I found many more bugs in the text, and more ideas for better presentation. It obviously helped that the audience today knew their way round the FreeBSD source tree much better. No Linux people, of course, though I did see somebody running Red Hat Linux on his laptop. I was more surprised that there were no NetBSD people there at all, and only one person who used OpenBSD, though he didn't do kernel debugging on it. That did make it easier to target things, and Martin Blapp came up with a crash in Netgraph which I was able to analyse successfully, much better than my final example (the one that caused so much pain at the end of last month).

In the evening to a reception. Realized that I had run into information overload when Mike Karels came up and said hello, and I didn't recognize him. Last year it happened with Bob Bruce. Talked to Mike and Donn Seely for a while: seems that Wind River is dropping BSD/OS. Sad times.


Wednesday, 10 September 2003 San Mateo
Top of page
previous day
next day
last day

First day of the main BSDCon conference. Met up with Ollivier and friends at breakfast, which took an inordinate amount of time. Decided to give ourselves a full half hour tomorrow. This is not my favourite hotel.

The keynote was Michi Henning doing his talk on progress in computing. I've heard it before, I thought in Brisbane at the Linux.conf.au, but he tells me that that talk was different, and that I must have heard it at an earlier AUUG conference.

After the break, some talks on SMP, devd and the new FreeBSD scheduler; it's sad how few non-FreeBSD people are present. Out to lunch at an Iranian (“Persian”) restaurant with Greg Sutter and a couple of Hennings (Michi, Thomas-Henning von Kamptz). Greg had to leave early to chair a session; somehow eating is taking longer than the time allotted. Caught a taxi back, only the second time in 21 years that I've caught a taxi in Silicon Valley (the first time was on Sunday). Listened to Luke Mewburn's talk on build.sh, then a longer presentation about long-range wireless, unfortunately less relevant to my interests than I thought. They did have a nice wireless receiver, though, which could be a basis for something to do at home.

In the evening a conference dinner, after which some project status reports. Interesting enough, I suppose.


Thursday, 11 September 2003 San Mateo Images for 11 September 2003
Top of page
previous day
next day
last day

In to breakfast earlier this morning, giving ourselves a full half hour. It didn't help: people just took more time. Ended up complaining about the poor service.

Missed the beginning of Poul-Henning Kamp's talk about GBDE as a result, after which Sam Leffler talked about his encryption work; it's refreshing to see how much attention he's paying to performance. Didn't listen much to Kirk's talk on UFS2: as he said on Monday, it's a watered down version of his walk through the code.

After the break, Peter Neumann talked about reliability, which was fascinating; for me, it was the best talk of the conference. In many ways, it ties in with what Michi Henning talked about yesterday. I'm astounded how many things he mentioned which I should have known but didn't.

The afternoon sessions were no less interesting: a group of Japanese students who did a userland BSD implementation somewhat reminiscent of User Mode Linux, and later John Baldwin showed how the Weather Channel uses FreeBSD for their rendering boxes.

Off after that with Ollivier Robert, Greg Sutter, John-Mark Gurney and Doug White to finish off the six-pack of Coopers' Sparkling Ale left over from the Speakers' Drinks that I had brought with me, after which off to half-heartedly hang around various BoFs, including a PGP key signing in which I didn't participate. Greg Sutter had brought a bottle of an interesting California wine (Cabernet Sauvignon/Merlot/Cabernet Franc) which we had intended to drink along with two bottles of wine from the Speakers' Drinks, but it got too late, we got too tired, and Greg disappeared in the critical moment, so off to bed instead.


Friday, 12 September 2003 San Mateo
Top of page
previous day
next day
last day

In for breakfast at 8:15 this morning to be able to be ready by 9 am; unfortunately, in vain. The waitress who “served“ us had obviously decided that we were not worth serving because we had complained yesterday. After 15 minutes, some of us had not been served, and despite my complaints, my food didn't arrive until 8:40, by which time we were involved in a heated dispute with a waitress who stated that she had more important things to do than to serve us. Marriott's obviously have a serious personnel issue.

The morning's sessions were interesting enough, but there were other things to finish off, and I was only half-listening. Managed to get everything finished by lunch time, and off to San Mateo for lunch.

After that with Ollivier Robert and Thomas-Henning von Kamptz to buy an Apple laptop for Thomas-Henning (in a record 21 minutes), then off with them to Fry's in Palo Alto, where I've never been before. Certainly a very different place from the one in Sunnyvale. I was looking for USB/serial converters and 802.11g wireless cards. The cheapest wireless cards with the Atheros chipset were $100, which sounds ridiculous in view of the fact that I paid $80 for the routers (802.11g/5 port Ethernet/much software) last weekend. Gave that a miss, and decided the USB converters were too expensive too, so just bought some relatively cheap NiMH rechargeable batteries and some batteries for my camera.

Then back to the hotel, where more people were still around than I thought. People were trying to work out how to tunnel out of the pay wireless Internet system (which, at $11 a day, was really pretty expensive). Wes Peters and Gordon Tetlow showed up from San Diego, and later off with them to Sam Leffler's place in the Berkeley Hills, and thence into Berkeley to have dinner in a Thai restaurant which I had been to before a few years ago.

To the hotel in Richmond, passing a satellite dish farm which could well be the uplink site for my satellite connection back home. To the “Quality Inn”, an interesting euphemism.


Saturday, 13 September 2003 San Mateo –> Richmond Images for 13 September 2003
Top of page
previous day
next day
last day

A busy night: the people next door appeared to be partying all night. When I banged on the connecting door, they called the security guards. sigh

After breakfast to the FreeBSD developer summit at Vicor. There were about 30 of us there, including notably Matt Dillon. Looks like we may be working to a good modus operandi. Much ground was covered; there will be summaries elsewhere.

In the evening to “Old Mac's”, a local restaurant of (justified) repute. Had one of the best racks of lamb I've ever had. Home early.


Sunday, 14 September 2003 Richmond –> San Francisco –> Los Andeles Images for 14 September 2003
Top of page
previous day
next day
last day

At least the people next door had checked out of the hotel, so the only noise was from the bikies who had arrived in the evening and departed loudly round 2:30 am. When I finally awoke from troubled sleep, discovered that I had fared better than Poul-Henning Kamp, who had been placed next door to a very active prostitute.

Things were less organized today—deliberately—but it gave Poul-Henning time to explain GEOM to us, much to my pleasure. Another real advantage was Warner Losh talking about debugging over firewire. I didn't realize that it was already available; the advantages are marvelous. Now I just need to get it to work at home, and then I can document it. A remarkably successful morning.

Things started to peter out in the early afternoon. I had gone to some trouble to coordinate transporters and transported to the airport—myself in the latter category—and had organized a couple of Dougs (Barton, White) to transport us to the afternoon. It's interesting to see how so many otherwise experienced international travelers get to the airport hours in advance. Poul-Henning was taking off 15 minutes before me, but wanted to leave an hour before to be sure of getting there. His justification was his lack of trust of the US system. Accepted that, but then at the very last minute Warner, who had previously not been involved, changed everything, so in the end he and Poul-Henning left with Doug B, and I left with Doug W an hour later.

At one point I almost thought Poul-Henning was right: I had my worst checkin ever in San Francisco, an airport I don't like anyway, and with which I've had trouble in the past. As every time, I had to submit myself to a “random” baggage X-ray. In view of the wine I had in the baggage (yes, we didn't get round to drinking it), I watched the security guard go through my baggage and investigate all of the supposedly 28 daemon statues (one turned out to be empty). Then to the security check, where it turned out that

Back to the checkin to check in my pouch, where I discovered that my ticket was missing. It turned out that the security check people had removed it from my bag and not replaced it. Even getting back there was difficult: I had great difficulty explaining to the people there that my ticket was missing, since I had a boarding pass with the flight coupon stapled to it. It seems that it surpassed their understanding that I might be going further than a single hop. It's fortunate that I had it, though: otherwise I wouldn't have been allowed far enough to get my ticket back.

Finally checked in my pouch, then back again for the third degree check, during which my laptop obligingly produced what they call a “code red”. Had to turn it on (fortunately they didn't wait for the “Windows” display), and they put it through their chemical testing thing again, and finally I was allowed through. Total time: 50 minutes. The way it was done doesn't exactly increase my faith in the methods. In particular, it should be a complete no-no for anything to be left behind when a passenger departs, and a chemical testing system that returns false positives (on a daily basis, according to one of the security personnel) sounds very dubious. It's worth noting that despite this delay, I was still well within time, and had an hour to spare in the lounge.

In Los Angeles, off to the Tom Bradley International terminal, where I was sent right back to terminal 4: my itinerary was wrong. I had to check in there again, since the agent in San Francisco hadn't been able to find my reservation. That doesn't explain why he checked my baggage through on the wrong trans-Pacific flight, though. When I finally did get my boarding pass, it didn't have a seat number on it. I had to go back to the gate and wait in a long queue to get that. Total time: 60 minutes, for an activity that shouldn't even have been needed.

Once on the plane, things didn't stop. Another baggage removal. As usual, it took 25 minutes, but it had to happen twice. As the captain reported, the computer systems in Los Angeles don't track these things adequately. What a day! I'll be interested to see if my baggage makes it to Sydney.


Monday, 15 September 2003 Images for 15 September 2003
Top of page
previous day
next day
last day

Never happened.


Tuesday, 16 September 2003 –> Sydney –> Echunga Images for 16 September 2003
Top of page
previous day
next day
last day

Arrived in Sydney and transferred to Adelaide without incident. Being home is certainly a change: it was raining when I left, I didn't see any rain during the trip, and it was raining when I got back. Things are very moist.

At home, finally got around to looking at the remains of echunga, which died ten days ago. Berndt was right: the capacitors had burst:


This should be exploding-capacitors-2.jpeg.  Is it missing?
Image title: exploding capacitors 2          Dimensions:          2048 x 1536, 560 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 Tuesday, 16 September 2003, thumbnails          All images taken on Tuesday, 16 September 2003, small
Diary entry for Tuesday, 16 September 2003 Complete exposure details

 

The capacitors are the tall green cylinders. Note particularly that they are bulging a little, and that there's a brown coating of electrolyte on the top of some of them. Typically this is caused by the use of inferior components, or components not rated for the job.

After that, went looking again at the dead wantadilla motherboard and found:


This should be exploding-capacitors-1.jpeg.  Is it missing?
Image title: exploding capacitors 1          Dimensions:          2048 x 1536, 592 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 Tuesday, 16 September 2003, thumbnails          All images taken on Tuesday, 16 September 2003, small
Diary entry for Tuesday, 16 September 2003 Complete exposure details

 

Both of these motherboards are made by Epox. echunga was just out of warranty. wantadilla was much less than two years old when it died. Those are the only two Epox motherboards I have ever bought. They'll stay that way.

I've since heard that, like many other manufacturers, Epox have been the victim of some kind of hoax with poor-quality electrolytic capacitors. That's possible, and I can sympathize. However, I've only had this problem with Epox, and so I'm still unlikely to buy their boards again in the near future.

Finished assembling one of my test boxes as the new echunga. I need to think about whether I really need two machines for my desktop; one might make things a lot easier. In the process, ended up with a total of four display cards in echunga, though currently I don't have a monitor for the fourth one. If I do go down to having only one machine, I'll certainly need to do some thinking of where I put it and how I connect it to all the monitors.

After that, setting up the Airlink+ wireless routers. The good news: they work as routers and access points, and roaming from one to the other works without any recognizable problems. The bad news: the configuration software is completely broken. It's impossible to change the IP address of the configuration address. Any attempt to change it results in a delay of 35 seconds or so, after which it remains unchanged. Similar things apply to an attempt to set the timekeeping (hard-wired to UTC+9, which is as close as it would get anyway: it doesn't allow for +9½h), and it certainly doesn't seem to do NTP. Checked the web site, but they don't have any upgraded software. Presumably they'll go broke before bringing out an upgrade.


Wednesday, 17 September 2003 Echunga
Top of page
previous day
next day
last day

Quiet day; didn't even catch up on mail. Instead did things like cooking, brewing and going to the dentist (I lost a filling at the BSDCon dinner). Also spent some time reshaping my network. I suppose it's a sign of the times that I can halve my Internet traffic by limiting the number of addresses visible to the network. Currently it's a Class C (192.109.197.0/24). By limiting it to 192.109.197.128/28, I can completely block all the spurious breakin traffic that makes up more than half of my total downlink traffic:

11:27:38.007025 sm200d < 217.81.33.99.14414 > 192.109.197.192.http: R 2103517920:2103517920(0) win 16384
11:27:38.450807 sm200d < 217.81.33.99.14517 > 192.109.197.192.http: S 2103517919:2103517919(0) win 16384 <mss 1400,nop,nop,sackOK> (DF)
11:27:39.666757 sm200d < 217.81.33.99.14517 > 192.109.197.192.http: R 2103517920:2103517920(0) win 16384
11:27:40.168923 sm200d < 212.253.9.84.4161 > 192.109.197.58.http: S 4247693221:4247693221(0) win 8760 <mss 1460,nop,nop,sackOK> (DF)
11:27:40.253626 sm200d < 217.6.224.162.2258 > 192.109.197.133.http: S 3810039922:3810039922(0) win 16384 <mss 1460,nop,nop,sackOK> (DF)
11:27:40.289702 sm200d < 217.81.33.99.14603 > 192.109.197.192.http: S 2103517919:2103517919(0) win 16384 <mss 1400,nop,nop,sackOK> (DF)
11:27:41.310236 sm200d < 219.165.96.2.26558 > 192.109.197.26.http: S 3406542070:3406542070(0) win 16384 <mss 1414,nop,nop,sackOK> (DF)
11:27:41.310315 sm200d < 217.81.33.99.14603 > 192.109.197.192.http: R 2103517920:2103517920(0) win 16384
11:27:43.395697 sm200d < 65.216.120.167.3904 > 192.109.197.189.http: S 1875904588:1875904588(0) win 16384 <mss 1460,nop,nop,sackOK> (DF)
11:27:43.443340 sm200d < 212.253.9.84.4161 > 192.109.197.58.http: S 4247693221:4247693221(0) win 8760 <mss 1460,nop,nop,sackOK> (DF)

That requires a lot of reallocation of IP addresses, of course, and NFS is causing problems; it accepts mount requests from alias addresses, but responds from the main address:

11:32:20.021709 192.109.197.142.980 > 192.109.197.135.111: udp 56
11:32:20.021936 192.109.197.80.111 > 192.109.197.142.980: udp 28

For the first time ever, I have kept a machine (battunga) running for more than 12 months, including an X session:

=== root@battunga (/dev/ttyp0) ~ 21 -> uptime
11:33AM  up 371 days, 12:59, 11 users, load averages: 0.05, 0.08, 0.06
=== root@battunga (/dev/ttyp0) ~ 22 -> ps aux | grep X
root      987  2.1 17.6 64948 28064  ??  S    12Sep02 1504:24.28 X :0 -bpp 16 (XFree86)

As a result, I don't want to do any rebooting. Things may look a little funny for a while.


Thursday, 18 September 2003 Echunga
Top of page
previous day
next day
last day

More catching up today. It's surprising how even minor changes in IP addresses can have far-reaching consequences. Apart from the problems with NFS, DNS also caused problems, and a lot of mail didn't get delivered. Years ago I set the default TTL for my zones to 86400 seconds (one day), but now that seems to be very long, especially since the network-visible name servers are off site.

Also spent some time playing with DHCP, and discovered that I had been using the wrong parameters all the time. Nevertheless, even the experts on IRC weren't able to explain why dhcpd rejected some requests and allocated an address out of the dynamic pool. Here be dragons.

As if that wasn't enough, for some reason ssh stopped working, and while I was investigating it, the satellite downlink died for a couple of hours. Not a productive day.


Friday, 19 September 2003 Echunga
Top of page
previous day
next day
last day

Yet another quiet day. Somehow I can't get up enough energy to do anything useful. It wasn't made any more easy by the fact that I need to write a book review for AUUGN about Jeff Duntemann's Drive-By Wi-Fi Guide. It turns out that writing book reviews is not easy, especially if the book is as bad as this one, and I spent lots of time trying to work out more diplomatic ways of saying “the name says it all: one person's view, superficial, buzzwords”. Didn't succeed.

Yet another Microsoft “virus”. This one claims to be a security update; possibly it is: if it takes the box off the Net altogether, it's probably doing everybody a service. The trouble is, it's big, and it's causing a lot of traffic.


Saturday, 20 September 2003 Echunga
Top of page
previous day
next day
last day

Somehow I'm not getting anything done. When writing even short documents, I seem to go through a period where just getting a basic structure in place is so painful that I take forever to do it. The last couple of days have been like that: I just need to write this damned book review and the president's column for AUUGN, but I can't get it done.

That's despite the fact that the president's column, in particular, is just opinion (something I certainly don't lack), and the topic, Internet abuse, seems good. When I came into the office this morning, I had 106 “Microsoft Security Update” messages, totalling 15 MB, waiting for me. Spent some time ensuring that that wouldn't happen again. The quick solution was a one-line change to /usr/local/share/spamassassin/50_scores.cf:

--- 50_scores.cf        2003/09/18 00:38:00     1.1
+++ 50_scores.cf        2003/09/20 00:28:56
@@ -968,7 +968,7 @@

 # highly generic tests for viruses that are scored just high enough to run

-score MICROSOFT_EXECUTABLE 0.100
+score MICROSOFT_EXECUTABLE 10.0
 score MIME_SUSPECT_NAME 0.100

 # GA never changes the whitelist/blacklist scores

This also has the advantage of stopping any other Microsoft executables, of course. The longer-term solution should be to reject them before they become too much of a traffic issue.


Sunday, 21 September 2003 Echunga
Top of page
previous day
next day
last day

Spring is finally showing signs of arriving, and the particularly heavy rain we've been having in the past few weeks has let up a bit, so off to Kuitpo Forest for the first time in months. About time I did some more relaxing things.

Apart from that, finally got my President's Column for AUUGN finished. I wish I could do these things faster. The book review still has to be reworked, as does another paper for the Singapore Computer Society next month, and also the column for Daemon News, now entering its seventh year. After that, I can finally get back to the Vinum work I'm planning to do.

These silly “Microsoft Security Updates” are still going around. My network traffic has more than doubled in the last couple of days. Tried a different approach to the problem, getting postfix to reject MIME headers with certain types. Upgraded to version 2 of postfix in the process, then added the following (courtesy of Perry Metzger) to my main.cf:

header_checks = regexp:/usr/local/etc/postfix/header_checks
header_checks contains:
/^Content-(Type|Disposition):.*(file)?name=.*\.(asd|bat|chm|cmd|com|dll|exe|hlp|hta|js|jse|lnk|ocx|pif|scr|shb|shm|shs|vb|vbe|vbs|vbx|vxd|wsf|wsh)/     REJECT We do not accept .${3} file types.  Please see http://www.lemis.com/email.html for further details.

Yes, that's all on one line.

Spring is finally showing signs of arriving, and the particularly heavy rain we've been having in the past few weeks has let up a bit, so off to Kuitpo Forest for the first time in months. About time I did some more relaxing things.

Apart from that, finally got my President's Column for AUUGN finished. I wish I could do these things faster. The book review still has to be reworked, as does another paper for the Singapore Computer Society next month, and also the column for Daemon News, now entering its seventh year. After that, I can finally get back to the Vinum work I'm planning to do.


Monday, 22 September 2003 Echunga
Top of page
previous day
next day
last day

Things are getting back closer to normal now, and I actually got some work done on Vinum. With any luck, and thanks to a slip in the schedule for FreeBSD 4.9, I'll be able to get it in that release. Also had a long and constructive discussion with Andy Johnstone of the SA Democrats. They're planning to reintroduce their Open Source Bill in parliament in the near future, and it's making a lot of progress since its last version.


Tuesday, 23 September 2003 Echunga
Top of page
previous day
next day
last day

Early on this morning I started rebuilding echunga with the new motherboard. It should have taken an hour or so; it took all day. Yet again everything went wrong. I always seem to have problems setting up these MSI boards: with the default configuration it showed my CPU (Athlon 1700 XP+) to be an 1100 MHz component. Based on my previous experience, went into the BIOS, which had set the FSB to 100 MHz (should have been 133 MHz). Tried to change that: according to the BIOS, I should be able to set it anywhere between 100 MHz and 280 MHz simply by typing in the value. Typing in values didn't work, and PgUp/PgDn limited me to 132 MHz maximum. Even then, the CPU was still reported with 1100 MHz, but it didn't find all its memory. Tried setting the CPU multiplier explicitly, with the result that the machine wouldn't power up at all. Checked the manual for helpful hints, but this is obviously something they didn't envisage: the “Troubleshooting” section answers questions like “What do you mean by PCB Version 1?”. Recovering messed up BIOS settings is something they want done via the BIOS setup screen. Finally found a URL for a boot recovery page; it no longer existed, and when I found the replacement, it still expected the machine to be runnable.

Finally reset the CMOS memory, which didn't work either. While I was at it, discovered a 100/133 MHz jumper next to the reset jumper, presumably what I should have set in the first place (last month's mother board didn't have a jumper like this). Wondered if the BIOS (which is obviously defective) had fried the CPU, so put it back into the old motherboard and confirmed that it's still functional. Put it back in the new motherboard and... it worked. Grrr.

Put the machine together and discovered to my surprise that it both had an on-board LAN connection, and that FreeBSD recognized it.

vr0: <VIA VT6102 Rhine II 10/100BaseTX> port 0xd400-0xd4ff mem 0xdfffad00-0xdfffadff irq 11 at device 18.0 on pci0
vr0: Ethernet address: 00:0c:76:40:43:ec
miibus1: <MII bus> on vr0
ukphy0: <Generic IEEE 802.3u media interface> on miibus1
ukphy0:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto

Getting the display boards to work wasn't that easy, though: Daniel O'Connor had given me a new (well, old) S3 PCI card which proved not to want to work with the others, so ended up with only four display cards in the machine.

As a result of the rebuild, and the lack of availability of Slot A motherboards, was left with a 750 MHz slot A Athlon processor (obviously), but also 512 MB of DDR memory, so put that in wantadilla, increasing its total memory to 1 GB. Noted in passing that I had made the swap partition exactly 1 GB, so I can no longer save processor dumps. One more sector would have made the difference. Grrr.

The 9 pm dumps didn't work. echunga's Ethernet interface (vr0, the VIA Ethernet interface, had hung itself up), and I had to reboot. Started the dumps again and went to bed. What a day!


Wednesday, 24 September 2003 Echunga
Top of page
previous day
next day
last day

Got into the office this morning and found that echunga's Ethernet interface had hung again. Gave up in disgust and put in another interface card, which fixed that problem. In the process, also finally migrated my primary X interface from echunga to wantadilla, which makes a lot more sense. In the process discovered that the version of x2x on wantadilla was 1.27, whereas echunga had 1.28beta1. Spent a lot of time fruitlessly looking for the source tarball for 1.28beta1 on the net (not found), then on tape (found, on a DDS-4 tape; I no longer have a functional drive). Shortly before desperation, discovered that I had, for some obscure reason, put it in my home directory.

This latest network worm is a real nuisance. Normally my network downlink traffic is about 15 to 30 MB per day; in the last couple of days it has gone up to about 70 MB, in danger of exceeding my monthly limit of 2 GB. And nearly all of it junk! My postfix rules ensure that these “security updates” don't make it here, but they don't stop the traffic. Spent some time working on that, and with the help of some rules published by an anonymous “Mark” managed to get a configuration file which rejects messages more often before hitting the data phase. It's difficult to tell whether it's making much difference, though.

In the afternoon to the dentists for a checkup, and then down to Grumpy's Brewhouse with a bottle of beer brewed last month, which tasted strange. They confirmed my suspicions: somehow it got infected, probably with an unpronounceable bacterium. Looks like I'll either have to drink it all now or pour it away. I suspect that our home water supply, though excellent for human consumption, may contain the odd impurity which ruins the beer. For the following brews I'll boil all the water first and see if that makes a difference.

Another power failure today. Now all our computers are either laptops or they're connected to UPSs, so nothing failed. Well, one of the wireless routers rebooted, but who cares? It turned out that Yana had reason to care: the router came back up in a default configuration, including a DHCP server which stole her IP address and cut her off from the net. These Airlink+ wireless routers must have some of the most broken firmware I've ever seen. Since it's also impossible to change their configuration server IP address, they both have the same one, and I need to physically disconnect one to configure the other. Ugh.


Thursday, 25 September 2003 Echunga
Top of page
previous day
next day
last day

Didn't do much “real” work today. In the morning, attending to a brew of beer that I started yesterday, and then bottled last week's brew; at the same time, also started some cooking for this evening, which unfortunately didn't turn out quite the way I was expecting.

In the evening, one of my spam blocking filters backfired on me:

/^From: .*Mail Delivery System/                 Discard Custom Header Virus rejection (From) (W32.Swen.A@mm)notification Rule 7815

This resulted in a mail loop which caused about 25,000 messages to get backed up. This recent mail worm is really the worst I've ever seen; I'm now exceeding my monthly Internet data quota just with this broken stuff. And I'm not even “at risk”; Microsoft users must have a much harder time than I do. It's time somebody started a class action suit against Microsoft.


Friday, 26 September 2003 Echunga
Top of page
previous day
next day
last day

Finally got back to work on Vinum today, a little late: today (US time) is also the date for the release of the FreeBSD 4.9 release candidate. It looks like there are still enough differences between releases 4 and 5 to make incorporation in 4.9 impossible at this short notice.


Saturday, 27 September 2003 Echunga
Top of page
previous day
next day
last day

Finally a bit of time to do some work. Gave up on getting Vinum up to date for revision 4.9: the deadline for the release candidate was today, and it was becoming clear that I couldn't test things properly in the time available. Spent some time doing other things instead.

In the afternoon, out riding in Kuitpo Forest again. The whole winter seems to have gone by without me finding time to ride more than once a month or so. Today at least we went a fair distance. The Peter Creek Road part of Kuitpo Forest is particularly pretty.


Sunday, 28 September 2003 Echunga
Top of page
previous day
next day
last day

It looks as if my attempts to limit mail have had some success. Today I got only about 43 MB of data:

0090BC014DB0,19.154565811157226,23.628793716430672,card1

That's still far too much, of course. Dan Shearer sent me some output indicating that exim can do better than this by first checking if the sender exists. It seems that the worm in question spoofs addresses, so this should be pretty effective. It would also be a good reason to change to exim.

Not a particularly active day. I'm getting ready for a lot of Vinum coding, if only I can find time after all the documentation I still need to write.


Monday, 29 September 2003 Echunga
Top of page
previous day
next day
last day

Finally got a bit of time to work on a number of technical things:


Tuesday, 30 September 2003 Echunga
Top of page
previous day

Spent the morning writing my article for Daemon News, the 31st in the series. Times certainly have changed in the last five years.

In the afternoon, more work on firewire debugging, and finally got it working... only to discover that the same problem exists as with serial debugging: I can't continue after entering the debugger. That's a real pain. Thought it might be a good idea to use the memory image facility: it should be possible to look at the memory of a crashed processor through the firewire interface. Unfortunately, I wasn't able to get that to work.


Top of page Previous month Greg's home page Today's diary entry Next month Greg's photos Copyright information

Valid XHTML 1.0!

$Id: diary-sep2003.php,v 1.49 2015/09/15 01:50:57 grog Exp $