|
| Greg's diary |
| April 2004 |
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Thursday, 1 April 2004 | Echunga | |
| Top of page | ||
| next day | ||
| last day |
Andrew Leppard along today to talk to me about the software we're developing. Despite (or maybe because of) all the documentation I've read, he was able to explain to me the way it works in relatively short time, though we spent a lot more time talking about unrelated things; he learnt the clarinet at school, and after a long hiatus is now getting back to playing it again, rather ambitiously starting off with Weber's first clarinet concerto. It seems that we have a number of people in the company who are interested in classical music.
| Friday, 2 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Mainly repetitive work today, getting the software to work locally, which it does well. Converting formats is less work than I initially thought, and it's definitely worthwhile: after a week of frustration with the “wrong” tools, it's making much more sense now.
| Saturday, 3 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Quiet day, spent reading and watching TV. In the evening, Max and Fiona Mitchell from Olivaylle came to spend the night: they're in a competition near Mount Barker tomorrow. Did another iteration of my Choucroute Garnie recipe. It's getting better, but resulted in a number of changes to the text anyway.
| Sunday, 4 April 2004 | Echunga | Images for 4 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
My beard is sold! But I only got $5.50 for it, so it doesn't sound like a way to increase family income to any significant degree.
Off to Jim Johnson's place for a barbecue today, and to meet both people working for Rocksoft and also a few people in the periphery. The weather was wonderful, the food was wonderful, and of course we had a lot of interesting discussions. I wasn't the only one who took a lot of photos.
Somehow, though, barbecues break the day. Back home and didn't do much beyond watching TV.
| Monday, 5 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent some time thinking about performance today. We don't seem to have done enough analysis of what the subsystem is doing in the way of I/O, so spent some time looking at that, and produced a surprising amount of debug code (about 1500 lines) in a surprisingly short time. It even worked without SIGSEGVing, but it didn't capture a number of system calls. I'll have to look at the linking in more detail.
Our Digitrex DVD recorder hung itself up again. This is not acceptable.
| Tuesday, 6 April 2004 | Echunga | Images for 6 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Today my new Brother HL-2700CN colour laser printer finally arrived, after a wait of nearly two weeks. It had interested me because it came with an optional duplex unit (costing nearly as much as the printer itself, unfortunately). Installation took its time, particularly because of the terrible documentation, but the printer itself worked out of the box. I suppose it's typical of modern equipment: it had a network connection, and it sufficed for me to plug it into the network for it to go out and find a DHCP server and configure itself:
Apr 6 10:56:22 echunga dhcpd: DHCPDISCOVER from 00:80:77:48:10:9c via xl0
Apr 6 10:56:23 echunga dhcpd: DHCPOFFER on 192.109.197.98 to 00:80:77:48:10:9c (BRN_48109C) via xl0
Apr 6 10:56:24 echunga dhcpd: if IN A BRN_48109C.lemis.com rrset doesn't exist add 43200 IN A BRN_48109C.lemis.com 192.109.197.98: timed out.
The last appears to be a dynamic DNS request.
Setting up the printcap entry was correspondingly easy:
lp|Brother HL-2700CN networked printer:\
:rm=lp:sd=/var/spool/output/lp:lf=/var/log/lpd-errs:mx#0:
All it needed was a DNS entry for the printer and a queue directory to make things work.
Installing the duplex unit was a different matter. Apart from pretty terrible instructions, I was able to attach it without too much trouble: this photo might help:
|
|||||||||||||||||||||||
After installation, though, it still printed on one side only. The documentation said nothing about having to do anything special, and there was nothing on the web site's FAQ about the problem. Just before calling technical support, I booted up one of my laptops with the Microsoft software that Dell forced me to buy with the laptop, and installed the management software. There, finally, I found a setting "DUPLEX ON/OFF", set to OFF. Set that, and it worked. While I was there, discovered at the same time all the protocols that the printer can handle. Quite a usable little program; it's really frustrating that the vendors have to supply it only for one operating system.
Things still aren't complete: for some reason, PostScript laser printers have a surprisingly long timeout on the last data, especially considering that PostScript tells you when it's done. On the final occasion, it took the printer something like 5 minutes to give up on the data and print the final page. I've found a couple of timeout parameters in the Microsoft-based program, but the help is non-existent (“no help found for this item”). It'll need some more playing around.
A number of people hate opinion surveys, but I like them. I feel that anybody who's as opinionated as I am should jump at the opportunity if somebody is actually interested in hearing these opinions. As a result, I've signed up on an opinion survey web site, Lightspeed. Today I discovered a mail message asking me to participate in a survey on food, something that greatly interests me.
When I tried to follow the link, though, I got:
Checking for scripting capability, and browser/operating system......
If you are not diverted in the next 10 seconds then your browser is not able to
support this survey - this may be because you have scripting disabled on your browser,
your browser is very old, you are not using a Windows operating system, or your
browser is not Netscape or Microsoft Internet Explorer.
In other words, this survey is only open to users of Microsoft systems. To be sure, I checked the page I had received (shown reformatted for legibility):
n=navigator;
u=n.userAgent;
if((u.indexOf('Win')!=-1
||u.indexOf('Windows')!=-1)
&& parseInt(n.appVersion)>3
&&(n.appName=='Netscape'
||((n.appName=='Microsoft Internet Explorer')
&&u.indexOf('Opera')==-1)))
{
window.location.href='../scripts/mrwebpl.dll?start&' +
'panelID=a543&surveyID=18&batchNo=4&project=B3qmw' + '&id=a543';
}
No question: this survey is only available to Microsoft users. How can an operator claim objectivity under those circumstances? I wonder if this doesn't offend against the Australian Trade Practices Act. And yes, I sent mail asking for an explanation, and of course got no reply.
| Wednesday, 7 April 2004 | Echunga | Images for 7 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent a bit of time this morning looking for yet more water processing equipment. Down to Meadows to talk to Bob Hay, the engineer, who'll process some copper pipe into a wort cooler, and to Mount Barker to look for a water filter for the domestic water supply. That proved to be very complicated, but finally ended up with a disk filter very similar to the one that we use for garden water, just with a 90 micron filter limit. That'll keep out the grit that clogs up the water system, but it won't keep out the crud. I'd need a second fine filter for that, but before I bother I'll see how necessary it is.
Apart from that, more investigation of my profiling code. It's interesting to note that I wrote the code in one day, and minor issues are keeping me busy. It seems that the C library has two symbols for each system call, for example open and __sys_open. It's not clear why, since they're both the same address, but it makes things very difficult. Found out that the loader has a facility for wrapping system calls:
`--wrap SYMBOL'
Use a wrapper function for SYMBOL. Any undefined reference to
SYMBOL will be resolved to `__wrap_SYMBOL'. Any undefined
reference to `__real_SYMBOL' will be resolved to SYMBOL.
This can be used to provide a wrapper for a system function. The
wrapper function should be called `__wrap_SYMBOL'. If it wishes
to call the system function, it should call `__real_SYMBOL'.
That looks like being a possibility, but it'll have to wait until tomorrow.
My printer is now happily printing double-sided documents—even if they're parts of different jobs from different systems. They don't seem to have thought that one through, and I certainly can't find anything in the documentation. If I don't find an answer soon, I should file a problem report.
| Thursday, 8 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent most of the day working on the profiling code, and ended up choosing a different way to handle things. Discovered in passing that the system defines not two, but three different symbols for each system call, for example open, _open and __sys_open. Reading the loader documentation pointed to a different option, --defsym, which creates additional symbolic names. Ended up creating a library by loading a relocatable object and then archiving it:
gcc -O2 -g -Wall -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -DDEBUG -I ../lib_rocksoft_c/ -Wall -c -o profile.o profile.c
ld profile.o -o libprofile.o -r --defsym _close=close \
--defsym _dup=dup --defsym _dup2=dup2 --defsym _fchdir=fchdir \
--defsym _fchflags=fchflags --defsym _fchmod=fchmod --defsym \
_fchown=fchown --defsym _fcntl=fcntl --defsym _flock=flock --defsym \
_fstat=fstat --defsym _fstatfs=fstatfs --defsym _pipe=pipe --defsym \
_pread=pread --defsym _pwrite=pwrite --defsym _read=read --defsym \
_readv=readv --defsym _write=write --defsym _writev=writev --defsym \
_access=access --defsym _chdir=chdir --defsym _chflags=chflags \
--defsym _chmod=chmod --defsym _chown=chown \
--defsym _eaccess=eaccess --defsym _getfsstat=getfsstat --defsym \
_lchflags=lchflags --defsym _lchmod=lchmod --defsym _lchown=lchown \
--defsym _link=link --defsym _lstat=lstat --defsym _mkdir=mkdir \
--defsym _mknod=mknod --defsym _open=open --defsym _readlink=readlink \
--defsym _rename=rename --defsym _rmdir=rmdir --defsym _stat=stat \
--defsym _statfs=statfs --defsym _symlink=symlink --defsym _sync=sync \
--defsym _unlink=unlink --defsym __sys_close=close --defsym \
__sys_dup=dup --defsym __sys_dup2=dup2 --defsym __sys_fchdir=fchdir \
--defsym __sys_fchflags=fchflags --defsym __sys_fchmod=fchmod --defsym \
__sys_fchown=fchown --defsym __sys_fcntl=fcntl --defsym \
__sys_flock=flock --defsym __sys_fstat=fstat --defsym \
__sys_fstatfs=fstatfs --defsym __sys_pipe=pipe --defsym \
__sys_pread=pread --defsym __sys_pwrite=pwrite --defsym \
__sys_read=read --defsym __sys_readv=readv --defsym __sys_write=write \
--defsym __sys_writev=writev --defsym __sys_access=access --defsym \
__sys_chdir=chdir --defsym __sys_chflags=chflags --defsym \
__sys_chmod=chmod --defsym __sys_chown=chown --defsym \
__sys_eaccess=eaccess --defsym \
__sys_getfsstat=getfsstat --defsym __sys_lchflags=lchflags --defsym \
__sys_lchmod=lchmod --defsym __sys_lchown=lchown --defsym \
__sys_link=link --defsym __sys_lstat=lstat --defsym __sys_mkdir=mkdir \
--defsym __sys_mknod=mknod --defsym __sys_open=open --defsym \
__sys_readlink=readlink --defsym __sys_rename=rename --defsym \
__sys_rmdir=rmdir --defsym __sys_stat=stat --defsym \
__sys_statfs=statfs --defsym __sys_symlink=symlink --defsym \
__sys_sync=sync --defsym __sys_unlink=unlink
ar sru libprofile.a libprofile.o
There should be an easier way.
With this finally working, found out some interesting facts. This program handles a plethora of files, but the profiling code only treats individual file descriptors, which get reused all the time. It looks as if I should track individual files, which will require yet another rewrite.
| Friday, 9 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Finally got a reply from Lightspeed:
Unfortunately, some surveys require certain operating systems, whereas,
others do not.
Based on the feedback received from you and other panellists, we will
start informing our panellists in the survey invitations and screens if
a certain operating system or software is required to participate in
that survey.
They don't seem to understand how unprofessional that is.
Apart from that, had a pretty quiet day. It's good to separate work from relaxation, something I haven't done too much of in the last few years.
| Saturday, 10 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Somehow all the TV web sites have conspired to produce no useful TV information today. I have a cron job that downloads information from eBroadcast, formerly Sofcom, but today the page was empty. The front page announced:
“The Australian TV Guide is currently offline during maintenance. Please check back soon. In the meantime, take a browse around eBroadcast Australia.”
“Maintenance” sounds to me like “crash”. I had already found that ABC, who used to have quite a detailed list, have succumbed to the web designer syndrome and have hidden all their information behind a maze of twisty little web pages, all broken. Is this a sign of the times, possibly derived from the Microsoft GUI paradigm, where showing too much information too easily is apparently frowned upon? Or do they just want to make it difficult for people (possibly including eBroadcast) to download the information easily? That would seem stupid, but then, the commercial channels Seven, Nine and Ten all do the same: badly laid out pages with minimal information, often combined with JavaScript to make it impossible to follow links, and which open tiny little non-resizeable windows without scroll bars, so small you can't read the bottom. Only SBS has a normal text programme, albeit with the rather strange option of sending you an SMS just before the programme begins.
The breakage is unbelievable. Spent some time investigating what it looked like under Microsoft. The answer: not as bad, though it was possible to completely lose the banner ads in the Seven Java popups, even with “Internet Explorer”. But in general, the breakage under Microsoft isn't as bad, and I'm left wondering which of the browsers is broken. Based on prior experience, it should be Microsoft, but I haven't found anything obvious that points in either way.
|
CUPS: “Open Source” gone mad
|
Topic: technology | Link here |
While on the subject of broken software, discussed on IRC the issues I still have with my Brother HL2700-CN colour laser printer:
The last page takes forever to print.
The duplex unit doesn't understand the concept of jobs. It's possible to print two one page jobs from different systems and have them both on one side of the same sheet of paper.
Somebody suggested that I should install CUPS. I had understood that it would be able to configure the printer, but this proved to be incorrect. Installing CUPS is a good way, however, to lose faith in the Ports Collection. I found:
The port didn't install the documentation.
After installing the documentation, which involved rewriting the Makefile, it turns out that there is no man page for cups(1) or cups(8) anyway. The closest is cupsd(8), the daemon, all of 35 lines long. There's nothing in the port installation process that tells you where to look.
After installing the HTML documentation, discovered that the index.html file reference non-existent files. It seems that it's intended to be accessed with a web browser via the daemon (http://localhost:631/).
Modern web browsers don't like the syntax http://localhost:631/: they want a fully-qualified name. So I put in the name of the system, and got an EPERM (permission denied) error.
With some further investigation, discovered that the configuration file /usr/local/etc/cups/cupsd.conf (referred to in the man pages as /etc/cups/cupsd.conf ) refuses even the local network, so it needed to be changed first. Not a big deal, except that there's nothing in the documentation to point to it.
Finally I was able to configure the printers with a very ordinary web interface. It didn't do anything (apart from overwrite my /etc/printcap) that lpd configuration couldn't do.
Trying to print proved that the support programs, such as lpr, hadn't been installed. It's not clear why, but after installing them (in /usr/local/bin/ and /usr/local/sbin/, they still didn't work properly. They also left the original lpd spooler files of the same name in /usr/bin/:
=== root@echunga (/dev/ttypf) /usr/local/etc/cups 234 -> wh lpr
-r-sr-sr-x 1 root daemon 27220 Dec 18 2002 /usr/bin/lpr
-rwxr-xr-x 1 root wheel 8520 Apr 10 15:02 /usr/local/bin/lpr
Depending on the sequence of pathnames in the PATH variable, this means that you'd get one or the other.
Round about this time I gave up. There's no reason to believe that CUPS can help me with the issues I have, and it's just too painful. The real disappointment is that it doesn't seem to do anything much that lpd doesn't already do.
As if that wasn't enough, my Digitrex DVD player hung itself up again. That's six times in 22 days. I'm still wondering what the cause is, but one way or the other it's completely unacceptable. Something will have to happen soon.
| Sunday, 11 April 2004 | Echunga | Images for 11 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Daniel O'Connor along today to bring some beer and play around with DVDs. Spent some time trying to do useful things: in particular, I was interested to find out whether the DVD burner in his Inspiron 8600 could burn one of my DVD+Rs, since all attempts in the Digitrex had failed. Spent quite some time trying to read in a pre-recorded DVD, during which Daniel discovered that his DVD reader had great difficulties with one of my pre-recorded DVD+RWs, though the Digitrex did not. Somehow the reliability of optical media is marginal at best.
In the meantime, installed mplayer on a number of machines, finally finding one on which it would work: it refused point blank on a system without audio hardware. Also discovered that you can mount a DVD as a normal file system. Here's a DVD+RW that I recorded
=== root@beeble (/dev/ttyp0) /wantadilla/home/grog 1 -> mount /cdrom=== root@beeble (/dev/ttyp0) /wantadilla/home/grog 2 -> l -R /cdrom/total 1 dr-xr-xr-x 1 root wheel 2048 Mar 22 08:05 video_rm dr-xr-xr-x 1 root wheel 2048 Mar 22 08:05 video_ts /cdrom/video_rm: total 1 -r-xr-xr-x 1 root wheel 116736 Mar 22 08:05 video_rm.bup -r-xr-xr-x 1 root wheel 8192 Mar 22 08:05 video_rm.dat -r-xr-xr-x 1 root wheel 116736 Mar 22 08:05 video_rm.ifo /cdrom/video_ts: total 8775 -r-xr-xr-x 1 root wheel 14336 Mar 22 08:05 video_ts.bup -r-xr-xr-x 1 root wheel 14336 Mar 22 08:05 video_ts.ifo -r-xr-xr-x 1 root wheel 65536 Mar 22 08:05 video_ts.vob -r-xr-xr-x 1 root wheel 102400 Mar 22 08:05 vts_01_0.bup -r-xr-xr-x 1 root wheel 102400 Mar 22 08:05 vts_01_0.ifo -r-xr-xr-x 1 root wheel 1073676288 Mar 22 08:05 vts_01_1.vob -r-xr-xr-x 1 root wheel 1073676288 Mar 22 08:05 vts_01_2.vob -r-xr-xr-x 1 root wheel 1073676288 Mar 22 08:05 vts_01_3.vob -r-xr-xr-x 1 root wheel 1073676288 Mar 22 08:05 vts_01_4.vob -r-xr-xr-x 1 root wheel 305528832 Mar 22 08:05 vts_01_5.vob -r-xr-xr-x 1 root wheel 100352 Mar 22 08:05 vts_02_0.bup -r-xr-xr-x 1 root wheel 100352 Mar 22 08:05 vts_02_0.ifo -r-xr-xr-x 1 root wheel 1073676288 Mar 22 08:05 vts_02_1.vob -r-xr-xr-x 1 root wheel 1073676288 Mar 22 08:05 vts_02_2.vob -r-xr-xr-x 1 root wheel 1073676288 Mar 22 08:05 vts_02_3.vob -r-xr-xr-x 1 root wheel 1073676288 Mar 22 08:05 vts_02_4.vob -r-xr-xr-x 1 root wheel 305528832 Mar 22 08:05 vts_02_5.vob
Even stranger is that you can play back the .vob files with mplayer, though I haven't yet understood the real implications. I must spend some time looking at the structures on disk. On the one hand the data is there and can be played, on the other hand I'm told it's encrypted, and the issues of region code are well known. But what's to stop somebody copying these files? I was able to copy them to disk with no problems.
| Monday, 12 April 2004 | Echunga | Images for 12 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another quiet day (a public holiday in Australia). Spent some time catching up on yesterday's events–the Digitrex hung itself up again–and considered alternatives. The fact that mplayer works relatively well made me consider downloading stuff from the TiVo, and found a version of mplayer which works with the TiVo, but didn't have the energy to try to work out where they had hidden the documentation. I think I'm building up to a big rant on how bad “Open Source” documentation is.
Our photocopier has been giving trouble. We had the problem a couple of years ago, where copies gradually got dirtier and dirtier. That was fixed when we cleaned it, but recently it ran out of toner and we had to replace the toner cartridge, after which the problem started again. Today I spent some time investigating it and discovered an undocumented used toner cartridge completely full with about 2 litres of used toner. I wonder if they cleaned it out last time. Getting rid of the toner was fun, as the photos show. Sadly, after cleaning, it wasn't much better. Back to the service shop.
| Tuesday, 13 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Back to work on program documentation today. I'm not sure whether I just take longer to understand programs, or whether I am just more insistent on correct understanding; in any case, in the present situation I am trying to find a more efficient way of doing things, so it's really important to understand the current implementation correctly first. It's heavy going: it seems that Ross' brain works differently from mine, and I'm spending a fair amount of time reformatting for different brain structures.
Got a couple of mail messages from people who thought I was being unfair to CUPS. Maybe they're correct, but both seem to have missed the point that it's the documentation that's the trouble, not the software. Specific issues were:
CUPS should be able to turn duplex on or off. This seems like a possibility, but since I can't find anything about it in the documentation, it's only theoretical.
Debian GNU/Linux should be able to do it better. This too sounds potentially possible, and I'll try it. But the writer goes on to say Install instructions are on the site, which suggests a similar problem exists there.
It's not a bug, it's a feature. There are no man pages because there's HTML documentation. This one seems to be based on incomplete reading of my rant: first, there are man pages, and secondly you can't access the man pages until you have read them.
I'm still left with the feeling that CUPS documentation is inadequate.
| Wednesday, 14 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent a little time this morning following up the claim that Debian installs CUPS better than FreeBSD. I was really expecting some improvement, but I didn't see it. I found exactly the same problems that I had had with FreeBSD. It did show that my accusations to the FreeBSD port were partially unjustified, though.
During the ensuing discussion on IRC, discovered some other interesting facts. The most important is probably that the printer has a complete web server installed in it, and that this web server appears to supply all the functionality of the maintenance program I talked about earlier, somewhat marred by the fact that some settings don't seem to show up correctly: although I was able to set a default Internet gateway, I couldn't get it to display correctly. Refreshing the display brought the home page, and returning to the TCP/IP config page showed a null gateway. Still, it's a far cry from the frustration of setting up serial printers 25 years ago.
In mail, some people on the UNIX heritage society were talking about John Lions' “Commentary on the Sixth Edition UNIX Operating System”, which had been leaked to the alt.folklore.computers newsgroup just under ten years ago. Surprisingly, it wasn't on the net anywhere, so dragged out the copy I made at the time and put it up on this web site. I must actually get round to reading it some time.
Also did some real work, more than this entry makes it look. Managed a low-hanging 7% performance improvement.
| Thursday, 15 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another day spent trying to document our software in the hope of understanding it better. It's certainly giving me more insights, but surprisingly little in the way of reportable results.
Heard from a number of people who have had problems with their Digitrex DVD recorder. I'm wondering if the things being sold on eBay are defective or rejects. Had a chat with the Australian Competition and Consumer Commission about the subject, and they took down a “complaint”, but it will only make sense if more people complain.
| Friday, 16 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another week seems to have gone by without too much to show for itself. Got a call from Greg Edmonds today, 11 months to the day since we got the results of our soil tests, saying that he had ordered some lime as fertilizer for the paddocks—after the lab report had clearly stated that we needed a lot of magnesium as well. This had been an issue last year: for some reason both John and Greg had been reluctant to even go looking for dolomite, and the result had been that we couldn't treat the soil before the autumn rains came.
Today spent some time looking for dolomite, and initially only found “dolomite sand”, of unknown composition. Yvonne went off to get some samples, along with some hydrochloric acid. She came back with two samples of dolomite sand and one of powdered dolomite (an equimolar mix of magnesium carbonate and calcium carbonate). The latter dissolved nicely in the acid, leaving a colourless solution; not very much of the dolomite sand dissolved, giving a green solution which later went yellow (so I suspect chromium, which is rather strange) and the remaining solid was black. I don't know what that has to do with dolomite.
The only problem was that the dolomite powder came in 50 kg sacks priced accordingly, and we needed 4 tonnes. It didn't take too long to find the right stuff, though, even deliverable tomorrow morning, at only half the price that John Westlake had quoted last year.
On the work front spent a remarkable amount of time discussing program documentation and (to a lesser extent) indentation style. It's amazing how excited people (myself included) can get about the subject.
| Saturday, 17 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
More mail on the subject of CUPS today. Although the correspondent didn't agree with my opinion of the CUPS documentation, he did point me to Eric Raymond, who does agree and has written his own account. His reasoning is very close to mine. A few days ago, I wrote: Installing CUPS is a good way, however, to lose faith in the Ports Collection. Eric writes: It has proved a textbook lesson in why nontechnical people run screaming from Unix. Eric had more stamina than I, and his document is well worth reading.
Somehow I had absolutely no motivation to do anything today. In the morning went to Mount Barker to buy some brewing equipment, and ended up spending a surprising amount of money on what appeared to be very few plumbing fittings, and as always was left with the feeling that what I had bought was not optimal. Still, it should be something for Bob Hay to put together next week. Had also intended to buy an aquarium pump for wort aeration, but came across a bewildering array of pumps and decided to ask some questions on the mailing lists first.
Not feeling much happier in the afternoon. Bottled some beer and started watching TV, only to discover that just about everything we had was on DVD, and the recorder was tied up all evening. I really must get sufficient motivation to set up a proper computer-based video system. I fear the biggest problem will be the user interface: if the big manufacturers haven't found one after decades of trying, what chance do I have?
| Sunday, 18 April 2004 | Echunga | Images for 18 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent the morning reading more of Eric Raymond's article on CUPS. There are some interesting parallels: he has written it only recently, and one of the printers he was trying to connect was a LaserJet 6MP. On the other hand, there are differences. Eric also used the web interface. To judge by the lack of commentary on the subject, it appears to have worked out of the box. But how? How did he solve the chicken and egg question at the beginning? In general the article goes into more detail than I would have liked, but he does come up with some good suggestions about how to document software.
Finally got round to riding again. We only have seven horses here, not enough to go out riding in the forest with Yvonne: two (Seren and Pebbles) are too old and lame to ride, two (El Mago Negro and Serafina) are too young, Guantanamera is too pregnant, and José is too timid to canter. Leaves only Darah, so I rode her in the arena in preparation for next month's gymkhana while Yvonne took many photos.
| Monday, 19 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another day spent with code analysis and documentation. There's an interesting tradeoff between basic functions and more comfortable ones built on top: the former are more difficult to use and (assuming a good implementation of the latter) more prone to error, but the latter are difficult to remember because of their sheer quantity. As functionality becomes increasingly complex, having such a large number of similar functions becomes more of a liability than an advantage.
In the afternoon to Bob Hay in Meadows, who has spent a couple of weeks preparing a wort cooler and a stockpot with a tap. The bill: a staggering $235. I had expected him to spend about an hour, but he said it was 4½. Admittedly, the work he did looks good, but I'm still baffled how he could take that long.
| Tuesday, 20 April 2004 | Echunga | Images for 20 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up early and into town to a board meeting of the IT Council of South Australia, the first one I managed to attend: last month I arrived 12 hours late. Presented AUUG, which evoked a certain interest. Afterwards talking about “Open Source” to Peter Griffith, chairman of the SA Chapter of ACS. I told him that I couldn't stand Microsoft “Windows XP“, because it was so difficult to find my way around a GUI. He agreed: he uses Microsoft NT 2000 “Server”. After thinking about it, it occurs to me that that's exactly the problem: GUIs are sufficiently non-intuitive that you have to learn them, and once you've learnt one, even a small difference is difficult.
Spent the rest of the day analysing code. I'm still not at all happy with the documentation.
| Wednesday, 21 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
We still haven't got through our discussions on code structure and documentation. Exchanged a grand total of 35 detailed mail messages on the subject, without coming to a complete agreement. The big issue at stake is: where and how do you document externally visible functions? My view is that there should be external documentation (typified by user guides and man pages, though man pages aren't necessarily the best example of such documentation), and that the commentary in the source code should be geared to the needs of the maintainer. As a result, the obvious place to describe how a function works is directly in front of the definition. For example, from Vinum:
/*
* Build up a request structure for writes.
* Return 0 if all subdisks involved in the request are up, 1 if some
* subdisks are not up, and -1 if the request is at least partially
* outside the bounds of the subdisks.
*/
enum requeststatus build_write_request (struct request *rq)
The others consider that the purpose of this documentation an externally visible function should be documented for the user, and that it should thus be in the header files, since that's the only part of the source that the user should be allowed to see. From my viewpoint, this is inconvenient because:
The user shouldn't be reading the header files just because he has them. He should be reading the documentation.
This makes two different rules for where to document a function, depending on whether they're externally visible or not.
The maintainer doesn't normally need to read header files unless he's changing the function itself, in which case he needs to get it in sync.
Most programming tools, such as ctags and etags, bring you more naturally to the function definition rather than to the declaration.
Having to look for things in two different places is just plain inconvenient.
None of this is a show-stopper, of course, but it's surprising how difficult it is to come to an agreement about it. Spent some time investigating alternatives to etags, and came up with GNU Global, which looks something like etags on steroids. It installed nicely and works pretty much like etags. I'll need to work out which of the additional functionality it provides is really of use; converting source files into hyperlinked web pages doesn't seem very useful to me, for example. One thing it doesn't do is to provide any easy way to get to the function declaration.
| Thursday, 22 April 2004 | Echunga | Images for 22 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Yesterday's exhausting discussion died suddenly, and we had no further exchanges today. Instead spent still more time analysing the code, and made a few changes to see what would happen. At some point I'm going to have to rewrite some of the code, so it's a good idea to get my feet wet with non-critical parts of the code.
| Friday, 23 April 2004 | Echunga | Images for 23 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Still more work on the code today, and came to the conclusion that I should be adding some improvements, which involved a 90 minute discussion of the matter with Ross. Looks like we've come up with some good ideas, which also deliver enhanced security “for free”. Spent most of the rest of the day considering the details.
Yvonne did some looking around for new wines today. In the past we've bought most of our wines directly from the vineyards in McLaren Vale, notably at Rosemount, but the wines are available for similar prices in local bottle shops, so she went out and looked, coming back only with a “white Shiraz”, really a rosé. Trying it brought home forcibly that we no longer needed to do what is standard practice in Germany: check whether it's sweet or dry. This Banrock Station wine was sugar water, and we poured it back into the bottle to use for cooking..
Note: I used to link to the McLaren Vale wineries page at Onkaparinga City web site, but they have become “modern” and hidden the link, forcing you to search for it with a broken interface. They also cater only for “Windows” users:
<META content="text/html; charset=windows-1252" http-equiv="Content-Type">^MIt requires JavaScript and doesn't work at all for text-based browsers. It makes assumptions about font sizes and doesn't work for the larger fonts needed for high-resolution displays. To add insult to injury, it insists on refreshing its pages every time you access them, and the W3 validation service finds 36 errors in the home page, after being helped with finding the undeclared DOCTYPE. Clearly I can't recommend this site any more. I may find a different link, but at the moment it's not worth the pain.
| Saturday, 24 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another quiet day. Weekends are beginning to be like that. Spent some time investigating computer-based TV systems. It seems that issue 8/2004 of c't, which I haven't received yet, has a number of articles on the subject, including a project based on Linux and the Hauppauge WinTV PVR 350, which seems to be the current favourite card. Spent some time looking for sources in Australia, which proved to be very expensive; in Singapore they're only half the price, so it may well make sense to have one sent here. It seems that there are still no FreeBSD drivers for the card, so there would be a bit of work involved.
Also considering film scanners; I have thousands of slides and negatives, though I wasn't able to find the ones that I was really looking for. Most better quality scanners now seem to have facilities for negative scanning, so it's really a matter of deciding which to buy. Sent out a message to a number of mailing lists. I wonder how they'll do with the colour negatives of the 60s, which are much more orange than nowadays.
| Sunday, 25 April 2004 | Anzac Day | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Yet another quiet day. After over 10 years chasing “Open Source” all day, every day, it's taking a while to come to terms with doing other things. It seems I'm not the only one: recently I've noticed a strong dip in the number of hits on my web site over the weekend. At first I thought it was something to do with bandwidth limitations on the host, but there's also a corresponding dip in the amount of mail we get. I should start doing some statistical analysis to ensure that I'm on the right track, but it seems as if the revolution is gradually tiring.
Spent today riding (only in the arena, practising for next week's gymkhana) and tidying up the HiFi cupboard, which was in need of it. Must also tidy up my office space; I can barely move in the Mike Smith Memorial Room. In the process managed to crash sat-gw, my downlink box, not once but twice: first I accidentally plugged a vacuum cleaner into the UPS, which promptly gave up the ghost, and the second time I managed to dislodge a power cable. That must be what happened to Yvonne when I was in Singapore last month.
| Monday, 26 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Public holiday for Anzac Day today, so again no normal work. Spent some (far too much) time hanging up pictures for Yvonne. I really need to get this place tidied up. Spent most of my time looking for tools, not all of which I found.
In the afternoon, thought it was about time to try out DragonFlyBSD, and downloaded an ISO snapshot. Installation was less than edifying: for whatever reason, the kernel didn't like the DVD drive from which it had been loaded, and I had to replace it with a CD-ROM drive. After that, the installation proceeded easily enough, considering there isn't any installation program. It does start a live system from the CD, though, along with an ssh daemon. ssh wants all kinds of security things, of course, and inside my firewalled local network I thought that I could risk telnet. Not so the dragonflies, it seems: there appears to be no telnet in the distribution.
Tried adding a normal user, somewhat difficult to do with a read-only file system. Ended up giving myself the home directory /tmp/grog, and with a bit of help from dhcp I was able to ssh into the system and do everything from my normal environment. Installation proceeded well enough, I thought, but at the end it didn't want to boot. More to look at tomorrow.
| Tuesday, 27 April 2004 | Echunga | Images for 27 April 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Back to work today, and wanted to get right into doing some performance testing. Despite a relatively minor change, which I expected to cause almost no improvement, performance went to hell in a handbasket, and eventually panicked the machine repeatedly; it looked like something to do with core dumping in a KSE environment. Spent some time looking at that and decided to install 4.10 Beta on one machine (zaphod, called daemon when it's running release 4), and to upgrade beeble to the latest -CURRENT.
Both took far longer than they should do. I've done this hundreds of times before; why should it be so difficult every time? People discussing things like total cost of ownership should be looking at these issues (not that it seems that commercial vendors are doing any better, but why should we measure things by them?). By the evening was able to reboot both machines, but it looks like I need Yet Another mergemaster run. sigh
Natalie, a friend of Yvonne's, arrived today. She'll be here for 10 days, so there'll be a lot of horsey stuff going on.
I've been advertising SPF records in my DNS for some time now, but yesterday Yvonne got a reject based on it:
<ers@mubo.de>: host ipev-mail.saar.de[192.109.53.24] said: 550 5.7.1
<yvonne@lemis.com>... Please see
http://spf.pobox.com/why.html?sender=yvonne%40lemis.com&ip=192.109.197.80&receiver=ip-comserv.saar.de
(in reply to MAIL FROM command)
I had set up my SPF TXT record using the “wizard” at http://spf.pobox.com/wizard.html, which told me:
|
||||||||||||||||||||||
On the other hand, the web error page said:
Why did SPF reject my mail?
Feb 6 2004
SPF is a standard extension to Internet email which protects people from email forgery.
ip-comserv.saar.de rejected a message claiming to be from yvonne@lemis.com.
ip-comserv.saar.de saw a message coming from the IP address 192.109.197.80 which is
blackwater.lemis.com; the sender claimed to be yvonne@lemis.com.
However, lemis.com has announced using SPF that it does not send mail out through
192.109.197.80. That is why the mail was rejected.
After a lot of investigation, discovered that the SPF wizard is at fault. It didn't catch a non-obvious error that I made: when setting address ranges, you should put them in the “IP networks” field (ip4:), not the a: field, which is reserved for server names. The wizard didn't notice that problem. The correct string should have been "v=spf1 mx ptr a:ozlabs.org ip4:192.109.197.0/24". Another indication that you should not put your trust in “wizards”.
| Wednesday, 28 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Yvonne and Natalie left early this morning to go to Olivaylle today, leaving me on my own for the day. Spent it trying to get my system upgrades completed, without success. Once upon a time it was as simple as make world, but nowadays every upgrade is a minefield. The FreeBSD project really needs to attend to this issue. It wasn't helped by a power failure in the middle of everything: I need to get some new UPSes.
Gave up on daemon, which kept SIGSEGVing login, and went back to running it as zaphod while I tried to coax beeble back to life: it had decided to return EPERM on any network packet. Got some work done, but not as much as I would have liked.
| Thursday, 29 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
More work on upgrading systems today, and finally got things up to date. It seems that the networking problems were due to a change in the startup procedures that didn't configure the firewall code correctly, so it ended up with only the one rule:
65535 0 0 deny ip from any to any
I shouldn't have had firewall code on that box anyway, of course, but it seems that configuration file changes that make such a difference should be avoidable.
Tried doing an “upgrade” install on daemon to bring it back to 4.9, so that I could restart the real upgrade, but after a successful upgrade it still SIGSEGVed on login, so ended up doing a cold install. This is really unacceptable.
| Friday, 30 April 2004 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Into town at lunch for another ADUUG lunch, this time at the Chianti Classico in Hutt Street. Nice restaurant, but a little too up-market for a geek lunch. Normally I drink beer at these lunches, but this was the sort of place where you would expect to have a multi-course meal and good wines. Might be worth returning to some evening.
After that, off to buy some hardware, including a Canon 9900F Scanner, with which I intend to scan in all my old photos.
After that to visit Peter Cassidy, who has a SPARCstation 5 which he can't boot. Proved to be an invalid ID PROM contents. We were able to boot from the disk, but of course we didn't know the root password. What a pain that even single-user mode requires a password. Noted also that software as recent as Solaris 2.5, on the disk, still starts a login with the erase character set to DEL. Considering that the main keyboard doesn't generate that character, this is really unbelievable. Took the machine with me; I'll play around with it for a while, though with only 32 MB of memory it might be better to cannibalize it and make one faster machine with flame.lemis.com, also a SPARCstation 5
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 | Previous month | Greg's home page | Today's diary entry | Next month | Greg's photos | Copyright information |