Older blog entries for chexum (starting at number 23)

last month
Nothing particular, I got distracted by interesting things. I wrote a replacement to nc, it had too many things done I did not wish. The result is yet another tcp proxy/tcp forwarder/etc, for a change, supporting connections through SOCKS5. openssh has a nice feature that allows it to use another program as a proxy connector, this is really a nice way to ssh out of a multi-level VPN-ed/firewalled network.

Spammers are still at it, I had to find a solution to refuse mail delivery to non-existent local users. The quirk is that I don't want to give up on qmail, which doesn't support this at all. A few days later, I'm using mailfront (with qmail still), which is an SMTP frontend to qmail. I prepared a few home-brewed patch (sent to the author), which does just that what I wanted. In turn, the author, Bruce Guenter, has an auto-reply bot to direct me to the mailing list with support questions... A nice idea with respect to the community knowledge, but I just sent a patch... Whatever, it's working well now, the rate of invalid recipients (forged by a Texan spammer) fell to a few tens per day. (Knock wood.) Some people still report abuse at our domain...

In this weekend, I planned to restore a not mission critical, but important Oracle server. After import failure, Oracle no longer starts. No help anywhere (and no DBA either, I just assumed that role...), but a lengthy internal error dump, which might have lead to the error, but nothing else since then. It just says mounted and opened, and in fact didn't. I went and went and recreated the database, then on Friday evening the server hiccuped (2.2.19 with ext3). Physical visit Saturday, kick some hardware, fsck in doubt to continue, bless some more, home again, recreating the database, importing the stuff again. A few hours later, another ext3 consistency failure. Great. No way I'm going to work more this weekend. On another server, I tried to update to 2.2.21, it didn't even come back up... That's all for work on Easter... :)

For a more pleasant surprise, I was always an avid fan of SVGATextMode, but it got stale, and the supply of VGA cards with supported chipsets dried up in the past years. Craving some fun, I almost accidentally tried the framebuffer support in the kernel, and while the first few experiences weren't of *anything* useful, it clicked together. Now I can use rivafb in 800x600@100Hz, with 37x88 consoles. Finally. I also updated my favourite console font (t.fnt, I can't live without it) to psfu, and 9x16 format (looked just wrong in 8x16), and now I finally have the same great console I had a few years ago, but with 100Hz, and working even better. To tell the truth, I was only playing with the fb stuff to see how the console would look like in 1024x768, in case I really get a TFT with a fixed resolution. Well, 1024x768 is still too large, but this fbcon stuff is great now...

It could spare me from a few months staring at 25x80 consoles, if people advertised fbcon as a real good replacement for SVGATextMode. I can't stand X and its fonts longer than about a half hour.

Another surprise is a Logitech Internet Keyboard. I just wanted to play with USB keyboards, and it looked great. Maybe toyish, but comfortably so. Well, it even feels great, better than the soft keys on the Amiga 1200, with a few (yet unused) extra toy keys, and a wheel, which works as a mouse wheel in X... If used with the PS/2 connector, this wheel is working as a cursor up/down wheel, another nice point. It's just that the Linux USB keyboard subsystem seems to translate the extra keycodes to something really different than in the PS/2 mode; even turns some of the buttons into *mouse* buttons... I always find rough edges in stuff. :)

I'm still toying with buying a: 15/17" TFT (probably not 17", it costs too much), some D-Link wireless cards (like scandal :) These DWL-650/500 things are everywhere apparently. I hope it's really available in our country, not just on price lists (has happened a few times). And finally an Ipaq/Jornada to run Linux, and have a wireless LCD terminal too. Not great plans for a life, I agree.. :)

rant again
I'm this >< close to giving up on email. The tons of spam on mailing lists is almost manageable, even with some Korean (and possibly some Chinese) spam, but now that Ralsky (spamford spawn) guy won't stop forging my hosts domain name. On top of that, clueless newbies engage in unsubscribe storms on every other mailing lists every week (and never reply to any help offer); quoting standards (and netiquette) fractured due to Microsoft MUAs; same company offers worldwide virus distribution services for anyone wanting it. The only (real) solution is throwaway addresses, which means giving up fighting spammers, and using their tricks.

TMDA? Yes, I was using a similar method almost five years ago, see <slrn5lrvt9.4dp$1@shadow.banki.hu>, but spammers still try to spam my addresses used in 1997...

OTOH, no other communication form is safe from these problems.. IRC, (any IM too, as I hear, I'm left out), web boards, SMS.

The Internet needs an overhaul as it stands, increasing capacity is no longer working (ignoring the fact that companies don't want to give it us either..) I've read a summary the other day about some traffic research in cities. The gist of it was if you increase the capacity of the roads, more people use them, and after a short period, it will be as crowded as now. Interestingly, it works both ways, if you reduce the capacity (road blocks, digging, parking restrictions), cars seem to disappear from the limited areas. The same seems to happen on internet. If you give a bunch of 2 hours@56k modem users 24h@10 mbits a day connection, they will use it up the same way doing nothing really more (but mirroring 600M iso images just to have them at hand). It must be our instinct... We must work it around, as much technically, as mentally. Is this the same as the tragedy of the commons?

mjg59: Thanks, that's the closest answer I was hoping, now I'm even more poised to play with these things. :) I've resisted impulse-buying until now. I'm not sure how long can I hold back..

May there be fun in everything..

Apparently raph captured the moment I was the last joined advogator. Everytime google puts up this page, I'm amazed at this timing :) Google finally believes that romfs is here. It took them about a month, and in the meantime, it sometimes mystically appeared as having content, sometimes as a default sourceforge page.

I'm also not particularly fond of "funny" release names, the fun wears out in nanoseconds.. There's also the thing with these newfound prereleases, release candidates, patch levels, builds.. Sometimes they have its place, but why is it so hard to give a release a version number? Why should Subversion 0.9 be called subversion-r1302.tar.gz? How could a dhcp-3.0rc8pl2.tar.gz, or a imirc-1.5-rc7.2.tar.gz happen at all? It's at least the same fun as Danish code names.. Man, I've never learnt English either, but I try to use it (hint to yosi :). And how about these myprogram.tar.gz, without any version marking? I'd be really glad if freshmeat didn't allow to post links to files like these (including myprogram-latest.tar.bz2, myprogram-current.tgz). Why would I want to inspect timestamps three days later to decide which was supposed to be the really latest? Call me grumpy, I'm done :)

System rebuild progressing. Most of the core system (including glibc 2.2.5) is recompiled with gcc 3.0.4 now. I found out the hard way that glibc doesn't like to be compiled with LFS default options (-D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64). gcc barfs along the way compiling pread.c, since that's using a system call with five arguments, one of it being a file offset, which, when turned into 64 bits, would need wider/more registers. First I thought it has something to do with --enable-threads, then suspected that C++ patch (to make artsd/KDE compile with gcc3). Fortunately an Athlon 1200 doesn't need that much time to compile anything. My first Linux machine, a 386DX40 with 8M, would take a few hours compiling the (then recent 1.1.3x) kernel.

I'm more and more addicted to switched ethernet. I also toyed with the idea of putting cables everywhere, "just in case", then with 802.11b to connect everyone in the neighborhood. Unfortunately, wireless access points don't have the flexibility of a Linux router, and putting a wireless card into a desktop PC (besides limiting it to places with worse connectivity) restricts it into connecting to one other access point, or to other cards using the same channel (while access points can communicate on many channels). IIRC. Is this correct? Time for more googling.

Finally, it occured to me the last missing piece everyone would need to connect the hi-fi and PC world, to overcome the ridiculously unprogrammable nature of hi-fi parts. What you would need is a dumb (maybe minimally manageable) box with an ethernet plug, and an S/PDIF output. This way, you could connect your DD/DTS receiver/amplifier to the network, and have hi-fi sound from your software without even any kind of sound card, and the PC (being an effective acoustic noise generator) could also be further from the listening place. Of course, you can then make sound from any of your PC's, and not run S/PDIF cables around, additionally you wouldn't need to buy an amplifier with multiple S/PDIF inputs... Of course, ethernet is not that reliable, but sound doesn't need more than 2 mbps, which is easy on 10mbps ethernet too.

Hm.. ethernet chips are cheap, ethernet transceivers are cheap, DRAM is cheap, S/PDIF is simple, code is cheap. Dream project. You heard it here first?

It would all be simpler if most stuff came with ethernet connectors. Think TVs, DVDs, phones. Nicer way to remote control, and/or route sound between them. Ideally, a phone should be just an interface to the telecom provider, a keypad, a mic and a speaker, and software "to bind them". Instead, we are getting handsets, message recorders, faxes, modems, voice modems, voip interface cards, ethernet phones that are just mic/speaker/keypad with voip ability, and more and more stuff doing the same thing just a bit differently.

Less fun today, maybe the weekend will be more fruitful; I'm starting to rebuild "my" distribution to the newest glibc/gcc/xfree et al. Since a week, I almost always wake up around 7:00 without any alarm set. Unfortunately, that's about half hour later which would be comfortable for breakfast and such.

We are preparing for some seat changes in the office (moving around real seats, not positions). Interesting times again.

Finally the machine is ready. I dug out an old CD-ROM, it's called GoldStar GCD-R520B. It didn't work. At least until I fixed the loose IDE cable. No problems since then. The drive has been manufacterd in June, 1995, and the vendor is now better known an LG, and they still have a web page about it. The drive is fine, that's what I call standards :) It can't read CDDA, but it's only a double speed drive anyway. Good for recovery boot, if I replace that stinky cable. Really surprising to have something this old, and mechanical still working nicely.

Maybe not at all, if I think about it. I also have an A3000UX which was in a previous life called cbmvie.commodore.co.at, until my friend bought it from the sell-out from Commodore. Unfortunately, most data from its hard disk is long gone, including traces of the Commodore System V Unix itself. I didn't know better. But at least it has a Commodore ethernet card, which is still a rarity.

Mostly installed the new machine. It took a while, because I didn't have a floppy for ages, I always install machines via hard disk copy, including lilo-ing them in the old machine. Most of the time this worked out ok, but now it didn't. First, the kernel failed to boot (reset after decompression). I tried to get rid of K7 optimizations, no go. Finally it occured to me that the APIC shouldn't be enabled with a K6. Then, at the early boot sequence, the new machine still crapped out with what looked like a corrupted filesystem, and lo, putting it back to the working one, indeed, it got corrupted. It took another while still I discovered that during boot, it set the IDE to UDMA/66 on a VIA controller which is capable only to UDMA/33... Majorly "restructuring" the boot filesystem in the progress.

But it's ok, by 01:35 (AM for you USians) I was ready, just forgot the BIOS/ATA settings in the most conservative state, and going to sleep early. Now, what day would be better for our remote sites to have urgent problems at 06:00 than this same day?

Thrown together a nice little spare machine (with bigger plans :) It's a K6-III (the cache everywhere thing). Now I just have to decide which hard disk I "spare" for it, and tomorrow I'll be getting a 100mb switch too to test my improved home network. Geeky. Gigabit would be better, have I mentioned it? :)

I've read a lot through ethernet docs these days, gigabit is very funny, it's full duplex on 4 wire pairs using all of them simultaneously in both directions. I'm not an EE, so it's magic to me :). You can't anymore confuse me with AUI, MAC, MII, GMII, PHY, PMD, PCS, MDI-X, MDI-II, ethernet flow control, link aggregation, vlan, NWay, LIT, NLP, FLP and such trivial things :)

Starting to envy the people who can toy with gigabit ethernet, I mean those who afford switches and anything else to connect it to. :)

Today is almost over, including a short session of a hopeful network redesign plans... It's a bit bizarre that this makes me happy :) 2.4.18-rc1 is out, 2.2 starts to look ages old on those servers..

14 older entries...

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!