Dakons blog

Rechner mit nur einem Betriebssystem sind zumindest in meinem Umfeld recht selten geworden. Seit einigen Jahren kann auch der Windows-Bootloader mit mehreren Windows-Installationen umgehen. Die freie Welt ist da natürlich schon seit vielen Jahren deutlich weiter, so bootet grub von nahezu beliebigen Medien was auch immer sich darauf zum Starten überreden lässt. Und die Installationsprogramme erkennen auch seit Ewigkeiten parallel installierte Windows- oder andere Systeme und richten dafür auch gleich einen Startmenüeintrag ein. Startmenü wie in "System starten", nicht wie in "Klicken Sie Start um zu Beenden".

Wie auch immer, in den letzten Tagen habe ich ein Testsystem installiert, auf dem neben einem XP32 und Win7-64 zwei openSuSE 11.2 ihren Platz fanden, nämlich jeweils ein mal in 32 und 64 Bit. Interessanter weise wird der Dual-Boot-Betrieb mit Windows automatisch eingerichtet und funktioniert ohne jede Beanstandung. Eine parallele Installation von 2 SuSEn ist jedoch nicht möglich: die Kerneldateien heißen für 32 und 64 Bit gleich und überschreiben sich gegenseitig...

Ich habe mir dann damit geholfen das ich dem 32 Bit-System schlicht einen selbstgebackenen Kernel 2.6.34 verpasst habe, das macht den Testraum auch etwas breiter. Aber peinlich finde ich das schon.

Recently I found a site with HTTP Content-Disposition header tests and their results. First that was only tested with Konqueror 3.5.8 from an ancient Knoppix CD, which the author quickly updated after I pointed him to the openSUSE 11.2 KDE Live CD. Nevertheless the results were still pretty bad, some things even got worse compared to 3.5.8. The attonlyucase and attwithasciifilenameucase tests are pretty stupid to fail.

So this is some basic technical stuff, includes collecting things from different RfCs, getting the implementation right etc., or to make a long story short: somethings that deserves my attraction ;) The other, simpler, but much more unlikely way to get my attention to such a problem would be some money (just in case *g*). Ok, anyway, the good news is: the current trunk is the best browser in this tests. Here is how the results would look like with the current trunk (previous results shamelessly stolen from the original site). I hope I can backport some or all of these fixes to 4.4.0 or 4.4.1.

Test Case		Konqueror 4.3.1	Konqueror trunk (r1075763)
Summary
Content-Disposition: Disposition-Type Inline	inlonly	pass	pass
	inlwithasciifilename	pass (filename information not used)	pass (filename information not always used)
	inlwithasciifilenamepdf	pass (filename information not used)	pass (filename information used)
Content-Disposition: Disposition-Type Attachment	attonly	pass	pass
	attonlyucase	fail	pass
	attwithasciifilename	pass	pass
	attwithasciifnescapedchar	pass	pass
	attwithfilenameandextparam	pass	pass
	attwithasciifilenameucase	fail (filename parameter is ignored)	pass
	attwithasciifilenamenq	warn (accepts the unquoted value)	warn (accepts the unquoted value)
	attwithisofnplain	pass	pass
	attwithutf8fnplain	pass	pass
	attwithfnrawpctenca	pass	pass
	attwithfnrawpctenclong	pass	pass
	attwithasciifilenamews1	pass	pass
	attwithasciifilenamews2	pass	pass
Content-Disposition: Additional Parameters	attcdate	unsupported (seems to ignore the parameter)	unsupported (seems to ignore the parameter)
	attmdate	unsupported (seems to ignore the parameter)	unsupported (seems to ignore the parameter)
Content-Disposition: Disposition-Type Extension	dispext	fail (does not treat it as 'attachment')	pass
RFC2231 Encoding: Character Sets	attwithisofn2231iso	unsupported	pass
	attwithfn2231utf8	unsupported	pass
	attwithfn2231noc	unsupported	pass
	attwithfn2231utf8comp	unsupported	pass
	attwithfn2231utf8-bad	unsupported	warn (displays the raw octet sequence as if it was ISO-8859-1)
	attwithfn2231ws1	unsupported	pass
	attwithfn2231ws2	unsupported	pass
	attwithfn2231ws3	unsupported	pass
	attwithfn2231quot	unsupported	pass
	attwithfn2231encmissing	unsupported	pass
RFC2231 Encoding: Continuations	attfncont	unsupported	pass
	attfncontenc	unsupported	pass
	attfncontlz	unsupported	pass
	attfncontnc	unsupported	pass
	attfnconts1	unsupported	pass
	attfncontord	unsupported	pass
RFC2231 Encoding: Fallback Behaviour	attfnboth	pass (picks the traditionally encoded value -- the one it understands)	pass (picks the RFC2231 encoded value)
	attfnboth2	pass (picks the traditionally encoded value -- the one it understands)	pass (picks the RFC2231 encoded value)
RFC2047 Encoding	attrfc2047token	fail (decodes it anyway to "foo-ä.html")	pass
	attrfc2047quoted	fail (decodes it anyway to "foo-ä.html")	pass

Update: the attwithfn2231noc testcase was added after I mentioned that particular unclarity.

Wie sonst soll man sich das erklären?

Last week I upgraded our companies Asterisk from 1.4 to 1.6. The thing that kept my busy for the most time was upgrading the fax application. To make things easier I try to write down what I found out.

First of all: app_rtxfax does not work with 1.6 anymore. Luckily app_fax now exists which is an internal module of the core asterisk package instead of an external module which makes things much easier in the future. You (still) need spandsp installed for it to work. A note for Gentoo users (like me in this case): don't forget to add the "span" use flag to net-misc/asterisk or you wont get it.

Ok, now you have an asterisk that can do what you want to but you have to tell him how. Time to change your dialplan! That part is rather simple: replace any "rxfax(foo)" by "ReceiveFAX(foo)" and any "txfax(foo|caller)" by "SendFAX(foo)". I had the additional problem that "System" did not work for me anymore to call my externals scripts so I changed that to AGI scripts, but that's probably my own fault because I did something wrong. Works anyway ;)

So, now your machine should be able to send and receive faxes again. To more problems may bite you: when your asterisk crashes every time you receive a fax you have triggered Asterisk bug 16374. Adding "-I" to the Asterisk commandline options like suggested in one comment helped for me. The other problem was that the page number in the emails generated by fax2mail script was always 0. The point is that my tiffinfo has "Page Number" instead of "Page" in it's output and the usage of "grep -c" there is simply wrong. My fix is:

PAGES=$(tiffinfo $SOURCEFILE | grep "^[ ]*Page Number:" | sed 's/.*:[ ]*//')

This is not another "new decade" post as you might expect. In fact, it's the opposite. I can once again go into nitpick mode and drive people nuts ;)

The first decade of the 21th century is not over. For the same reason that the 3rd millenium and the 21th century did not begin 10 years ago: the simply did not. A decade has 10 years. A century 100. A millenium has 1000. The first year was 1. Ask our calendar guys: there is no year 0 in the "christian calendars". See you next year! ;)

While I was hacking on improving KGpg status output when generating a key I found some behaviour in the plasma progress bar that first confused me (as I thought my messages were not working at all) and now I think it's just a bug.

As you can see right of the application icon there is no description shown for an already running job. Once it is finished the description is shown there. I found out that this wasn't my fault as the dolphin progress bar showed the same behaviour when I was copying some large files from my brothers computer.

I'm using KDE 4.3.3 (yes, I'm lame) and would appreciate any hints on if that's intended, if there's a bug report, or anything else that would allow me improving this. My mail address is in your KGpg about box ;)

I have two problems with a new feature in KGpg. The first: I thought there was a feature request for this but I can't find it anymore. Well, anyway, I've nearly done it. Still reading? Ok, then I'll tell you what the heck I'm talking about.

I assume you know how the web of trust works (if not, read about it). There are basically two views on the web of trust: the "absolute" view, i.e. the view using every key on the keyservers (like e.g. http://pgp.cs.uu.nl/ does or the view from inside your keyring. Now I wrote some code that just does that: it tries to find the shortest path from one key to another in your keyring. Currently the output looks like this:

max depth: 2
key 1D8EC918F4F358BCE08E4456009C42DBCC5CF182 depth 1
key 8B8B22090C6F7C47B1EAEE75D6B72EB1A7F1DB43 depth 0
key CAA9884B19C0E72D9D9023EB3B68865AD86A0D19 depth 1

This are all possibilities in my keyring to go from my key to our beloved Plasma-guy. Now I would like to get this into something more shiny. I already have a model that has all the information about the keys. I would love to get a helping hand to get this in a beautiful shape.

There are some simple rules: I only display shortest paths. Maybe something like maximum trust path will come up for KDE 4.5. The destination key is the only key at level 0, the source key is the only key at level n. Typical values for n will be something like 3-5. Every key at level x has at least one connection to a key on level x+1 and one to a key on level x-1 (the start and end keys of course have connections only in one direction). No key can have a connection to a different level as it's direct neighbours.

To get some idea I have two sets of keys that should be enough for testing while I beat the remaining code into shape. The first one is the example from above, every key is linked to every other key at the neighboring levels (i.e. 4 links).

The second one would have 7 keys: start is key A, at level 1 there are three keys (B, C, D). At level 2 there are 2 keys (E, F) each one with two connections to level 1 keys (E to B and C, F to B and D). Finally there is the destination key on level 3 and the obvious connections betwenn start and destination and their neighbours.

So if anyone with some GraphicsItem foo (or whatever fits best) would like to give me a helping hand: fire up your KGpg, there you will find my email address.

SSchäuble. Guttenberg. Von der Laien. Merkel. Eigentlich muss jeder hier wissen das die CDU absolut unwählbar ist. Nehmen wir den Wolfgang. Nach 100000 € von einem Waffenhändler und Rücktritt von allen Ämtern macht er jetzt auf Grundgesetzminister. Und wie! Oder die Laienministerin. Nach Herdprämie auch noch Internetzensur. Sie kann sich einfach nicht entscheiden ob sie die Gesellschaft an den Anfang des letzten Jahrhunderts oder nach 1984 zurückwerfen will. Es wird der Anfang sein, Zensur hat es da auch schon mal gegeben. Oder das Kanzler. Im Ausland macht sie immer auf gut Freund mit jedem, im Inland schafft sie es nicht mal ihren eigenen Ministern zu sagen wo es langgeht wie es scheint. Christian Wulff sagte er hätte nicht den absoluten Willen zur Macht und wolle nicht Kanzler werden. Merkel hat definitiv immer den Willen gehabt nach ganz oben zu kommen. Was sie dort machen will hat sie sich vorher wohl nicht überlegt.

Ach ja, bevor ich es vergesse. Internetkompetenz und so, is' klar. Wir haben die Kraft. Alles roger.

Der Regierung Schröder musste man am Ende zu Gute halten, dass sie Deutschland wenigstens weitgehend aus dem unsinnigen Irakkrieg heraus gehalten haben. Was man der SPD aus der derzeitigen Legislaturperiode positiv anrechnen kann ist mir auch nach zwei Wochen noch nicht eingefallen. Sie haben am Atomausstieg festgehalten, den die Grünen ursprünglich gefordert und durchgesetzt haben. Ein Abrücken davon hätte sie auch den letzten Rest der Simulation von Standhaftigkeit gekostet und einen der beiden einzigen Minister mit halbwegs positivem Stimmungsbild wohl völlig aus dem Rennen geworfen. Aber sonst? Man hat SSchäuble seine Vorratsdatenspeicherung und von der Laien ihre Internetzensur durchgewunken. Als man den Koalitionspartner bei letzterem einige unwesentliche Details hat abringen können hat man es diesem auch noch durchgehen lassen das er es als eigenen Verdienst verkaufen konnte. Und dann stellt man sich hin und behauptet etwas von Internetkompetenz und das man die Bürgerrechte schützen will? Unser Land kann mehr, das ich nicht lache. Es kann mehr, aber ihr nicht!