...doesn't have to be high tech. I've got a new toy that I really like: It's a Aristo 868 slide rule, and I find it pretty amazing how much you can do with it. Best of all: no batteries :-)

I'm obviously not against high tech, for example:

That's my trusty HP 28S, which I got in 1988 - RPN forever! But still there is a certain minimalistic appeal to the simple magic of sliding log scales...

It's a good day in my world when my favourite game, Sturmovik, fully works on Linux under Wine, and better, more reliably and at least as fast as on Windoof.

That also includes the TrackIR, thanks to these guys and their Linux-Track project - which is still a bit rough in places but generally works fine with a variety of hardware (from webcams to TrackIRs).

In addition to the Linux-track stuff you'll also want the Linux-Track WINE plugin which presents Windoof apps with a TrackIR-compatible API. That thing was a pain to get running, and you might want to check out this patch by me to make it work properly with recent Linux-Track revisions.

Apart from those: no real problems after some initial conf.ini tinkering; no more inexplicable stuttering under Windoof, no more dualbooting and smooth performance with graphics options close to the top levels.

One of the banks I deal with uses one-time transaction numbers which they send in bulk by paper mail every now and then. (Pity that the local banks aren't as enlightened...)

The notion of "TransAction Numbers" I like, carrying the paper slip I don't - because paper encrypts so very badly and I'm lugging my Palm with me all the time anyway.

gocr takes care of the OCR, and generally works fine but BSTS...if comparing two sheets of meaningless numbers wasn't so ridiculously, mind-numbingly, dull. Can't have that.

So I had to look for a cheap, quick and dirty solution for that not-quite-problem, and after ten minutes I had it: espeak.

It's a fairly simple speech synthesizer, which unfortunately insists on pronouncing numbers as numbers, not individual digits, but a trivial half-line of perl data massaging took care of that.

Sure, espeak sounds like a post-lobotomy HAL 9000 with a hangover, but hey, it makes sanity-checking of the OCR results a lot faster and easier.

(Enlightened) Laziness is a virtue :-)

The Citibank onscreen keyboard reappeared after an overhaul of their web site (as usual not making it any better, but Now With More Shiny Crap!). Anyway, it's 2010 and onscreen keyboards are still a damn stupid idea for password entry - and so my Citibank-Demouse Greasemonkey script got an update.

If you are using spamassassin without sa-update you will not like to hear that as of 5 days ago spamassassin has pretty much declared open season on all your mails:

Due to an incredibly gross and dirty bit of rule all properly dated mails get an extra 3.6 added to their score. *kablam*!

Botch/Fix: edit /usr/share/spamassassin/72_active.cf, and change the regexp for FH_DATE_PAST_20XX to something that doesn't fire in the near future (like 20[2-9][0-9]). Don't forget to sa-compile if you use compiled spamassassin rules.

This is not a Beatles song.

Horricle + Sun = the horror, the horror.

I've got a netvista thinclient serving as jukebox in my bedroom. Recently I wanted to switch to a faster wireless adapter, which required moving to the 2.6 kernel series. This is especially painful for the netvistas, but nevertheless possible (despite some sources claiming that it won't work). Here's the rundown.
(more...)

I've got a new toy,
(more...)

I hate udev. It does not work in settings very crucial and important to me (ldap-nss) and it's a huge step back from hotplug in terms of useful functionality. Stupid complicated config environment, bloated *and* it does not load modules on demand. Dear udev authors: you can keep that crap and i'll stick to what works, is small and almost semi-elegant: hotplug.
(more...)

Emacs and exmh go together very well, but of course there's spots where things rub across. Today I scratched such an itch successfully: I now have access to exmh's address database from emacs (and so can you).
(more...)

...but quite nice. I've got my own ideas about how to manage my collection of digital photos and so I wrote my own tiny, idiosyncratic but sufficient photo manager five years ago (says rcs; hmm, those years went quickly!).

Now that Conny has a shiny digital camera of her own (and a bit of associated trigger-happiness) she also needs something to organize her pics with. And while my photomanager is fine for me Old Fart, it's a little bit gnarly. So I looked at more user-friendly (but not idiot-friendly) solutions. And voila, the first apt-cache hit was already what I had been looking for.

Martin Herrmann has written "martin's picture viewer" aka mapivi, which is more than just a viewer (a feature which is fairly irrelevant to me). It's written in Perl plus Tk (important to me), it's a photo manager (ditto) and it keeps pretty much all info where relevant: in the photo files themselves. The last is most important IMHO, because it frees me from sundry databases, proprietary overview formats and the like. mapivi uses EXIF and IPTC metadata to record pretty much anything you can think of in extra segments of your jpegs (and other image formats that allow such metadata storage).

The thing is a bit rough in places but works very well for a 0.x release, and the combo of Perl and Tk is really fun to work with.

I've immediately gone full steam ahead and coded the two plugins I need to emulate the few features my photomanager had over mapivi (complete with balloon popup help texts for Conny); also submitted one patch to the upstream author.

Gone is my photomanager, and welcome mapivi. Not Invented Here indeed :-)

...with a FREE spelling mistake, but nevertheless excellent. (source:some flicker user via Cryptogram)

It's amazing what they managed to squeeze out of that poor 6502...

today i got my new (well, less than a decade old) sun to replace my last dead ultra2: a netra 120 (aka sun fire v120). but what a noisy bugger!
(more...)

should you be so unlucky to have to use a box with onboard Intel "Graphics", specifically a 965Q chipset, and also one of those fat Dell screens which suck at all resolutions but 1680x1050, and maybe even want to run debian Etch (without testing or experimental or somesuch), then you'll hate all the involved hardware manufacturers. (Likely you do that anyway, for unrelated crimes against syadmins but regardless: THIS time i did find a way to coerce them to function.)
(more...)

...but rather with them. Nevertheless, here's a pretty cool game I've started playing recently: Terminus, a space FPS/RPG thingie which rocks: a solid Newtonian physics model, good graphics, both linux and win supported natively (on the retail cds) and so on.

What's even better is that the Underdogs have the game downloadable in its full glory (abandonware; released in 2000).

...and it doesn't dig many if not most of my files. Damn dumb beast!

Well, no more. swish-e seems to be better behaved, and actually works! *duh*

These guys have cooked up a tiny perl CGI frontend (which I've reworked and cut down a lot further), and the search functionality on this site works again.

I've also fixed a long-standing annoyance of blosxom: plugins can't cleanly set the title of a page from the story title, because the header plugins run first and the story plugins have no official access to the output. The fix is Really Dirty, in the best tradition of blosxom which is Abysmally Dirty Code: a plugin with a sub last {...} that massages $blosxom::output. If it finds exactly one story in there, then it changes the <title> to that story's title. Hideous but it works, and the search interface can display story titles instead of just the boring story links.

If you want to play with the Abominable Code for this stuff, let me know. fakefake

I dislike spam, very much, and repeat offenders deserve all my wrath. Here's another use of the iptables recent module in a very cheap and simple manner, to limit the spam blasters' effects on me and my servers' life.

(I've said nice things about ipt_recent before here and here, both with example applications.)

I've just added these extra rules to the firewall setting on my mail servers:

 # smtp access is controlled by previous behaviour: spam me and you lose.
iptables -A INPUT -p tcp --dport 25 -j smtplimited
 # smtp: if mimedefang has flagged you as bad, you lose for 12h
iptables -A smtplimited -m recent --name SMTP --hitcount 1 \
   --seconds $((12*3600)) --rcheck -j TARPIT
 # clean up the old entries to unclog ipt_recent
iptables -A smtplimited -m recent --name SMTP --remove
 # and let people through if they've been good in the past
iptables -A smtplimited -j ACCEPT

My mimedefang filter has been instructed to (do the perl equivalent of) echo "+$ASSHOLE_IP" > /proc/net/ipt_recent/SMTP whenever it detects an asshole that tries to:

• send email with a non-existent sender's address
• send email to one of the spamtraps within my domains
• send email to nonexistent addresses within my domains
• send me spam (where the degree of spamminess is beyond any reason for doubt)
• send me viruses

(The decision logic is actually a bit more complicated: I certainly don't blacklist known forwarders and backup MXes.)

The net effect is that when you do something nasty to me (email-wise), all your subsequent connections to my mail servers are tarpitted for the next 12 hours. Works great, easy to tweak if you want to be more lenient (just up the hitcount and adjust the following --revove rule) and reduces the time my systems have to waste on repeating the checks for surefire rejections on the smtp-envelope level. (I usually get about 5000-10000 rejections per server per day.)

Damn, how I hate those wobdesigner idiots that make you click on virtual keyboards to enter your passwords. Virginmoney/Westpac has such a thing, and a very crappy one, too, but we have a fix. All this involves EczemaScript, not nice but better than suffering.

Now I've got another set of suckers to deal with, Citibank AU. Their setup is less gnarly but still annoying. This time, I produced a fix myself: Citibank-Demouse is a Greasemonkey script that simply clears the hooks that invoke the virtual keyboard; and hey presto! keyboard-entry of your password works again.

Federico Schwindt pointed out the appropriate Citibank UK url, and the script has been updated with this information.

(Clearly not.) But nevertheless, this is a very interesting article about Fitts' Law and how interface design should be affected.

Some of the assertions seem...unfitting, though: for me, moving the mouse to the bottom is the most annoying move, not the least: the mouse has to travel beneath my palm and wrist (I tend to control the mouse with my fingertips and a bend of my wrist and rarely ever lift my elbow off the desk.) Moving to the top I just extend my fingers, so that's faster.

But then I'm weird: I have the mouse on either side, with some bias to the left -- but I'm somewhere between righthanded and ambidextrous otherwise, I switch between two different keyboard layouts every day (german at home, english at work) and so on.

I complained about bank logins a number of times, the last one being citibank's mouse mess. Apart from that I have no major issues with their credit card service...except the insistence on a PDF plugin to read your statement. WTF?!

Well, now they can insist as much as they want because I wield the Greasy Monkey wrench...and I win! This greasemonkey script neuters the plugin-insisting code and also converts the EMBEDded (*spit!*) PDF into a normal plain link. Works for Citibank AU, maybe other incarnations. (BTW, userscripts.org sucks, I lost my password and can't reset it, thus this script will not end up there anytime soon.)

It's not exactly the CSS Zen Garden but not too bad either. I think.

Two weeks ago I rebuilt the chgc website from scratch, with nice new images, no more tables, standards compliant HTML and CSS and so on. I also got rid of \rho's HTML++ thingie and replaced the automation guts with Mason (but still statically rendering everything).

Comparing this with the current setup I'm pleased with the results.

Perl is a programming language, just like C, only it is even weirder.

$ perl -e '$a=3; $b=++$a + $a++; print "$b\n";'
9
$ perl -e '$a=3; $b=++$a + ++$a; print "$b\n";'
10
$ perl -e '$a=3; $b=$a-- + $a++; print "$b\n";'
5
$ perl -e '$a=3; $b=--$a + ++$a; print "$b\n";'
6

My Bizarrotron just broke its indicator needle. Fascinating!

I'm no fan of EczemaScript but the recent Westpac/Virgin mess convinced me that scripting in the browser is not necessarily bad - as long as I control what runs where and when (hence I use NoScript and deny all scripts everywhere except where I know the code).

Yesterday I mucked around with making the local TV guide website bearable - all I really want to see is the innermost table of actual information, minus all the square acres of blinking advertisements and similar drivel.

Upgraded to Firefox 1.5 (actually painless, very different from past experiences), installed the newest (0.6.4) Greasemonkey, found a script that claimed to fix that mess but which was too ugly by far, rewrote it to suit my prefs, done.

I learned a lot about Javascript and DOM (and also where Greasemonkey sucks) than I ever wanted to, but that's fine.

Today I thought about tackling the Virgin problem, but found out that Joel Hockey has already written a nifty simple small script that gives you text-based password entry back (without removing the silly buttons, should you be stupid enough to want them). Thanks, Joel!

But it didn't work. My stubbornness has few limits (and the weather was not flyable today), so I learned still more about JS and GM and the DOM, especially about the recent paranoia that badly affect the new Greasemonkey and wrecks most of the nice things about DOM and JS (if there were any in the first place).

In the end I fixed Joel's script (and sent it back upstream) and am now quite pleased with my army of greased monkeys.

Next step, maybe: adding a squad of platypuses.

I rebuilt my backup-and-music box recently, because the case was ugly and because I needed to fit another disk for online backup.

I had a Sun 811 case lying around, and another similar to a 411. Neither would take the Epia motherboard, PSU, dvd burner, two 3.5" disks and a 20x4 lcd, but together they can throw off the yoke of conformist PCism! *ahem*

So I cut out the plastic top of the 811 and riveted the 411 onto it, which gives me space for the drives. A face for the open rear end of the 411 was cut from the cannibalized pieces of my Sony stereo junk and hot-glued in. The frame for the HDs is an old cut-up drive bay, and the support for the burner is a piece of sheet metal that I riveted in (hot-glue isn't strong enough and I didn't want to use expoxy for no reason at all).

The front with the lcd got a painted fascia (balsa) and the IR sensor was mounted internally this time. After a shitload of further surgery on the cases and innards I ended up with this pleasant look. But you can't see the rear in that photo which is good. None of my small ATX power supplies would fit without totally rebuilding the thing (not-so-perfect an idea as I'd basically have to strip all insulation off, then resolder half the high and low voltage connections and cram all the resulting mess into the franken-case), so I started looking at DC-DC PSUs. Like this one. Which I did eventually buy, thinking "the 90W/145W peak PSU I have used so far, so this 200W thing should do nicely". Cost me about us$100 (with a 9A AC-DC external brick and shipping).

Little did I know, and for that matter, too little effort did I spend on research. Plug it in, fire up, works - somewhat: now I get loads of noise on the audio out connection. Not just mains hum but all kinds of activity-dependent crap as well. This is when I started doing the research I should have done before. It turns out that loads of people hate the PW-200-M for being a crap piece of equipment. First, it's nowhere near 200W, and some other speciality PSU manufacturers have accused the makers of shoddy lying advertising. The 5V rail sagged under the load of my two disks down to less than 3V at times. The 12V line is not regulated, so iff you're not using a regulated brick you'll fry your gear (especially the carputer people hate it for that). The smoothing caps are not exactly large at 390-1000uF. (But the form factor rocks, which is why I bought it...)

Tried pretty much everything non-destructive, like powering only the board from the PW-200-M, trying different 12V supplies to verify the noise is coming from the PW-200-M etc...but no joy. It may be useful for really low-power scenarios where one doesn't care so much about power quality (i.e. non-audio application), but for me it's junk...Bugger.

Back to square one: normal PSUs don't fit. Most high-quality DC-DC PSUs like the Opus gear won't fit or require 19V like the DC2DC converters.... So for the time being, I plopped my normal small ATX PSU like an outboard motor behind the box...with some shielding and extra grounding it doesn't affect radio reception too much. *sigh*

...you see an ad on your Coke bottle that says "5 HUGE GIGS" and you think nah, 5 gigs is nothing special today but I remember when it was not just huge but UNIMAGINABLE only to *blink* and realize that they were talking about *music*...

The iptables recent match module is pretty cool; things like keeping the sodding ssh brute force guessers at bay are trivial: accept only X new connections to the ssh port within a minute, if not coming from a trusted known network. Two iptables-lines.

Unfortunately, the module isn't overly stable internally and there's some rollover bugs like this one. I'd still give it some extra coolness points for allowing me to implement Port Knocking without any userland tools in 5 minutes:
(more...)

The expletive (in Austrian German) applies to the Linksys people, who managed to castrate the WAP11 access point in its version 2.8 but good:
(more...)

A French court has recently decreed that using P2P systems is legal if it's for private use. This includes both up- and downloads. Sweeeet! French Freedom Fries! A nice result that doesn't make the Content Cartel happy.

Link to an article (in German) about this

Applied to computers this kiddo-truism would be like: "Lapdogs are obsolete the moment they leave the factory, and extinct by the time you buy it. You want software to work on it? Here's your two rocks, bang them together and see how you go..."

I've been Having Fun with kernel 2.6.14 and my machines. Lots of Bloody Fun. It takes heaps longer to configure things. The documentation has not exactly gotten better. The (feature-)stability of the 2.6 series is a joke. Some things still don't work. Lots of new things have stopped working. WAAAAAAAH.

The lucky list: ide-cd and ide-scsi still conflict. The latter ist needed for reasonable cd burning. The module documentation blithely says "There is usually no reason to remove modules, but some buggy modules require it". Idiots. The xserver will make your box hiccup badly and fuck up playing of sound if you run it with the previously required niceness. Vmware modules don't build on 2.6 at all, but somebody has cooked up a (really ugly but working) patch. The devmapper maintainer is a clue-resistant idiot who repeatedly refused a one-liner fix for a problem that breaks the use of the disk group so I rolled my own packages. The maestro3 sound support has gotten worse, the chip gets confused every now and then now (and I'm not going with the ALSA suggestion: You can install that bloated crap when you pry the keyboard from my cold, dead fingers.) Loopback (ahem, devmapper) encryption is still not possible for non-root users. Wavemon does no longer work. The netfilter code is fucked up, IP_NF_NAT_LOCAL is gone since around 2.6.11 which means that natting local conns doesn't work anymore. My nice location-independent setup for the proxy (everything configured to use localhost:3128, then NAT that to the real proxy if needed) is now officially unsupported. Thank you, bastards!

And, of course, direct rendering for mach64-derivates is once again absolutely utterly fucked up (uncompilable, incompatible, non-working code). Might be a good thing that with trying to find out and fix all those niggling problems I've got no time to play any games anyway...

On the plus side, however, are things like the kernel key storage api: goodbye quintuple-agent, hello kernel! I'm currently experimenting with code to make that stuff easier to use; Debian packages to follow as soon as things stabilise...