Or, BEID of the same version. Oh well.
It took me a while, but I finally managed to jump through all the hoops[1] required by backports.org correctly and in the right order for the package of the above version to appear there. So if you're running sarge, have an electronic ID card and a smartcard reader, and want to see what the government knows about you, then go ahead and install them. More info on how that works can be found on the backports.org website.
There are two known issues as of now: first, you may get wrong root certificate errors; second, you need to install all packages from the belpic source package, or some things may fail. I intend to properly fix these two bugs Real Soon Now(TM), but I'm having issues reproducing the first one and am trying to wrap my head around being policy-compliant and produce working packages at the same time, which is required to fix the second one. It's slightly ugly; believe me, you don't want to hear the details.
Speaking of ugly, SCons is not ugly. It's downright hideous. I'll have autotools any day.
If you're not running sarge but are using Ubuntu 6.06 'Dapper Drake', then you will need different backports. Luc Stroobant was kind enough to provide them; use deb http://www.stroobant.be/eid ./ in your sources.list to get them.
[1] they are not many, but I managed to break my upload three times or so before it actually worked. And since I don't know about any status page or so, I needed to ask people every time—which I don't want to overdo, so I give them a while before investigating. Next time I probably won't...
Customization-modules
There's some discussion going on in Debian about firmware in main. Some people want it, others don't. The main argument by the firmware-in-main advocates seems to be that we can't install hardware if there isn't any support for it in the installer, and that the installer can only install by using stuff that is part of main. The obvious answer to that would be to modify the installer so that it would support retrieving stuff from non-free, but then that would require some changes to the installer which, according to Joey Hess, are rather significant.
So we now have a GR proposal to allow for firmware in main anyway, so that we can support non-free hardware. Obviously the proposal includes overriding the social contract. And huge and insane flame wars.
I don't like the sound of that. And I'm not convinced that the only way to get non-free udebs on a system is to download them from the Internet; Windows, RedHat, SuSE and other operating systems have supported the concept of "driver disks" for ages; I don't see why we couldn't do the same. And since our system is way more flexible than theirs anyway (it being modular and all), we can easily implement this using udebs.
So rather than engaging in flame wars, I just sat down and wrote some code that would allow the concept of "customization disks": CD-ROMs, hard disk partitions, USB keys, or floppy disks with udebs on them that will be installed if the user chooses the right option in the installation menu (requiring expert mode to get there, however). This can be used for all kinds of interesting things: non-free modules, but also OEMs (say, HP) providing udebs on their driver CD-ROMs so that they can have the installer support their hardware which is supported by the latest version of the Linux kernel, but occasionally not by the version that Debian Stable ships with, by backporting a kernel driver to the right version. Should they so choose. Or people could use it to test new modules without having to upload them and without having to set up their own repository. Or third-party developers could use it to add extra code to the installer to support some automagic stuff, whatever.
It hasn't been tested at all yet, but at this point it's not much more than a proof of concept anyway. Comments are welcome, but please read the thread on -boot on the subject, too.
Screamers
Yesterday evening, there was a movie on vt4 called Screamers. Pretty nice movie, that. Some science fiction in which man-made self-replicating AI weapons get out of control into a Frankenstein-esque apocalypse. It was made on a tight budget (which shows at times), but was overall quite good. Sometimes funny, but not silly. Sometimes having a romantic moment, but not so much that it gets meek. Sometimes scary, but not so much that you can't bear it (I once literally zapped away from looking at Sigourney Weaver hopping around spaceships in Alien because I was getting too nervous at what I knew would eventually happen—even before anything did, in fact, happen).
Most of all, the movie has a story to tell, and it's one that makes one think about AI: do we want it, or should we be afraid of what we might unleash? There's no answer, and there doesn't need to be one—but it's still a good question.
Web design
Proper web design seems to be something that is out of grasp of the majority of commercial web developers.
My dad asked me for help with some particular website last week, and in the two minutes that I was looking at the site, I saw four errors:
- It only worked if you use IE, giving you a white empty page if you use a real browser. I thought the dark ages of "This site optimized for Internet Explorer" were over three centuries ago.
- When opened in IE, it started off with a useless intro. I wasn't even remotely impressed, neither was my dad. Especially not seen the fact that he had to jump through a number of hoops to get at this silly intro (IE is hidden away into obscurity on my dad's computer, since I don't want to have to deal with the incredible amount of security issues this sieve has)
- The first page given was a menu which is never repeated afterwards (hurray for continuity), and which uses mystery meat navigation.
- Last but not least, <li> items (or whatever they abused to implement the same thing) have their first sentence coloured in a certain kind of blue, with no underlining. <a> items are coloured in the exact same shade of blue, with no underlining (and no hover effects). So the only way to see that a piece of text is no hyperlink is by hovering your mouse over there and verifying whether the mouse pointer changes. Way to go!
I forgot what the exact URL of the website was. It's not very important, though.
Re: Undelivered Mail Returned to Sender
Dear SMTP Server program at host orange.fr, It's great if you inform me that my message could not be delivered to one or more recipients. It's also wonderful if you attach it below, so that I can verify what's going on. It's less great if verification shows that the mail you're replying to was sent by me on 24 july 2005, which is over a year ago. It's not very interesting if verification also shows that the mails in question were actually mailing list posts, meaning that you sent the bounce to the wrong address (it should've gone to the mailing list bounce address instead). It's rather annoying if you repeat this behaviour sixteen times. I guess I should call me lucky that I only sent four mails on that day, and that only four people on your mailserver were subscribed that day. No, I did not contact postmaster, since the mails weren't all that important, and I doubt there's anything they can do about it at this point. I certainly don't need assistance with these mails. Also, if I would address your postmaster the way you suggest it to me, those mails would end up in my own mailbox anyway, rather than in your postmaster's mailbox. Your postmaster is, however, welcome to contact me if they so desire. Love, Wouter
Ska & Newwave
I have three machines located under my desk at home:
- rock, an SMP i386 machine on which I'm typing this,
- ska, an MVME167 machine that has a 68040 processor and is supposed to be a Debian/m68k buildd host,
- and newwave, a Macintosh Quadra 700, which does not have enough RAM to be used as a buildd, but which I use to do some m68k-based hacking on.
A few months ago, the last two had died with various problems that had seemed to be disk-related; both gave an error in the firmware—with ska's error obviously being more... detailed... than the 'unhappy mac' error icon that newwave gave me.
Today, I powered both on again. newwave didn't even give me a hiccup; it booted right away, even if it needed to do a rather extensive fsck run. ska did take a little more work, unfortunately; it seems to boot, but it does give me SCSI errors, too. At least it's managed to make it to a login prompt, however, so I can have a chance at fixing it.
Interestingly, just as I finished the above paragraph, and less than half a minute after it made it to said login prompt, it breaks down rather horribly:
scsi0 channel 0 : resetting for second half of retries.
SCSI bus is being reset for host 0 channel 0.
scsi0 : DCMD|DBC=0x98080000, DNAD=0x2cec6c (virt 0x002cec6c)
DSA=0x3ad300 (virt 0x003ad300)
DSPS=0x2060000, TEMP=0x3ad154 (virt 0x003ad154), DMODE=0xe0
SXFER=0x58, SCNTL3=0x0
phase=MSGIN, 0 bytes in SCSI FIFO
SCRATCH=0x3ad300, saved2_dsa=0x3ad300
scsi0 : DSP 0x2cec64 (virt 0x002cec64) ->
0x2cec64 (+1a1) : 0x98080000 0x02060000
0x2cec6c (+1a3) : 0xc0000004 0x002ce5c4 0xfff47010
0x2cec78 (+1a6) : 0x80080000 0x002cfb80
0x2cec80 (+1a8) : 0x98080000 0x00030000
0x2cec88 (+1aa) : 0x50000000 0x002cede0
0x2cec90 (+1ac) : 0x60000200 0x00000000
scsi0 : issue queue
scsi0 : dsa at phys 0x3ab188 (virt 0x003ab188)
+ 64 : dsa_msgout length = 2282225664, data = 0x2ce7d8 (virt 0x002ce7d8)
+ 60 : select_indirect = 0xfff47010
+ 56 : dsa_cmnd = 0x2ce5c4 result = 0x8888, target = 27147, lun = 0, cmd = VENDOR SPECIFIC(0xe5) c4 ff f4 70 10 88 08 00 00
+ 48 : dsa_next = 0x0
scsi0 target 27147 : sxfer_sanity = 0x0, scntl3_sanity = 0x0
script : 0x0 0x0 0x0 0x0 0x0 0x0 0x0 0x0
Unable to handle kernel NULL pointer dereference at virtual address 0000004a
Oops: 00000000
PC: [<000bfae2>]
SR: 2700 SP: 01ad9cd8 a2: 01ad8000
d0: 00000001 d1: 00002622 d2: 0000000a d3: 00000008
d4: 002d2704 d5: 002d14e0 a0: 00000000 a1: 0015aefc
Process ntp_time_as_com (pid: 775, stackpage=01ad9000)
Frame format=7 eff addr=01ad9d5c ssw=0505 faddr=0000004a
wb 1 stat/addr/data: 0005 003ab188 01ad9d60
wb 2 stat/addr/data: 0005 0016f8b5 0000abd6
wb 3 stat/addr/data: 0005 01ad9d40 0000000a
push data: 01ad9d60 0000adec 0015aee0 00000001
Stack from 01ad9d40:
0000000a 00000008 002d2704 002d14e0 002ce5c4 002ce140 003ab188 0000ac4a
000bfe68 002ce5c4 0013e781 00000000 00000018 002d2704 00000000 00000000
0000ac4a 003ab188 002d14e0 002ce000 000bfefc 002d14e0 003ab188 0013e576
00000000 0000ac4a 002d14e0 003ad300 98080000 000bf586 fff47000 002cec98
002ce000 000c02d4 002d14e0 003a5200 00000000 000b2704 00000018 00000020
00000000 00000000 002d14e0 002ce000 000bf912 002d14e0 002d14e0 00000001
Call Trace:
[<0000ac4a>] [<000bfe68>] [<0013e781>] [<0000ac4a>]
[<000bfefc>] [<0013e576>] [<0000ac4a>] [<000bf586>]
[<000c02d4>] [<000b2704>] [<000bf912>] [<000b6c4c>]
[<000b5f24>] [<000b6a04>] [<000b6c4c>] [<000b6c4c>]
[<000b65b6>] [<0013d5fe>] [<00002404>] [<000beb82>]
[<000bd206>] [<000bf022>] [<000bf4d6>] [<000bed46>]
[<000833fe>] [<00004e52>] [<00003b86>] [<0000c00c>]
Code: 2c68 004a 4878 0180 2f0c 47fa 0c8a 4e93 508f 2a0b
Kernel panic: Aiee, killing interrupt handler!
In interrupt handler - not syncing
aw.
Oh well, at least I've got one more working machine now. Only need to update it from unstable as it was 41 days ago to today's unstable. Going to take a while, I fear.
Finally,
Google Maps has reached civilization.
GR season
It seems to be the season to propose general resolutions. We seem to be having
- Proposals to decide what to do with non-free material that somehow isn't software (but is software after all).
- Proposals to decide what to do with our beloved DPL. Or not so beloved, apparently.
- Proposals to stop proposing proposals.
- And perhaps some non-finished and as of yet unproposed proposals, too. But I wouldn't know about that, oh no!
Personally, I've lost count as to how many proposals there are out now, and how many of them got the required amount of seconds and whatnot. Lucky me I'm not the secretary.
But know what, since it seems to be trendy to propose GR's these days, I'll add my contribution for today:
-----BEGIN PROPOSAL----- The DPL, seen as how he's fairly important for our inner workings, should not choose his job lightly; having a bad DPL might have a detrimental effect on the project as a whole. Therefore, we must make sure that we discourage our fellow developers from becoming DPL unless they mean it. It is, thus, imperative to make the DPL job not a very nice one. In that light, the Debian Project resolves that: When the DPL makes an important decision that will have an effect on the project as a whole, he is only allowed to do so provided he balances on his left hand, puts his right big toe in his left ear and has his left toe point towards the ceiling, types the command with his right hand, doing all this while poking his nose with his tongue. The help of a yoga master in training for this procedure is allowed; the help of a yoga master or anyone else while actually performing the procedure for real is not allowed. To prove that the DPL did in fact follow procedure, a video should be produced by the DPL or any of his relatives which is GPG-signed by at least the DPL and a second developer who witnessed the proceedings first-hand. -----END PROPOSAL-----
A nice side-effect of this is that there will be no more DPL decisions once this resolution passes. That seems to be what people want.
(oh, and for all you who take yourselves too seriously: the above is called "satire")
Command frequency
nice one-liner to find out what the highest number of commands is that you use. For me, the output is:
wouter@country:~$ history|awk '{print $2}'|awk 'BEGIN {FS="|"} {print $1}'|sort|uniq -c|sort -rn|head 10
98 ls
71 cd
25 mutt
24 sudo
18 ssh
18 killall
17 xine
17 apt-cache
11 less
10 man
Having ls be number one isn't something that surprises me, at all. I have this silly habit of typing "ls" in my terminal window when I forget what I was about to do. Obviously that doesn't help me remember, and sometimes I do it two or three times in the same terminal window without anything in between—as if anything would've changed...
However, having killall in the top ten is something that does surprise me. It appears I tend to get pretty impatient when things don't do what I expect them to...
Then again, the output of 'history' doesn't seem to be very complete, having only 500 entries...
Infodrom quotes
Nothing like reading quotes by yourself out of contet.
I think I prefer this one:
And "88" is based on 2 eights, which is 2 more than 6. This means that there are 3 numbers 2 in the number 88; with 8 being 2 more than 6, the number is actually 666, which is the number of the devil. Therefore, they *must* be associated with the Devil.
I can't fully remember what the context for this one was (nor can I for most of the other quotes on that page), and google doesn't help. I do have a faint idea that it involved some case where I thought someone else was basing an argument on something ridiculous, and that I wanted to show them that; but I'm not sure what it exactly was. Anyone else have an idea here?
GDM
Joey,
Debian's GDM does ship with the default file that contains lots and lots of comments. However, if you use the builtin configuration tool at least once, it will overwrite your configuration file with another file that contains just the configuration, not the comments anymore.
Perhaps that's what happened to you?
HP, you idiot.
The three ML110 systems that I ordered for a customer just arrived. The customer wanted RAID1, so I ordered three extra hard disks. The hardware bits that hold disks in its place are based on some ingenious slide-and-click system that does not require me to do some acrobatic funny things with a screwdriver just to be able to put a hard disk in the machine.
This is good.
The slide-and-click system requires me to put screws in the drive that have heads of a certain specific size (it uses that as the thing to lock against). Obviously all the industry-standard screws that I have are way too large to fit. The machine does not come with extra screws, nor does the disk.
This is not so good.
Update: 1s/HP/Wouter/. The servers do contain those extra screws, I'm just blind. Whoops.
Whoops.
20:10 < Yoe> how do I ask irssii to connect to a second network again? --- Log closed do sep 28 20:12:47 2006
Note to self: if you're going to ask whether the "/server" command does in fact disconnect oneself from the network, do not start a phrase with the "/server" keystroke...
heh.