08:36:52 | seletz | hi! |
09:31:06 | seletz | erikm: hi! |
09:31:18 | erikm | hello |
09:32:15 | seletz | erikm: I read on LAK that you did some work on the smc9196 driver code. Is that true? |
09:32:36 | erikm | no |
09:32:41 | seletz | hmm |
09:32:44 | erikm | I've been working on the cs89x0 |
09:32:44 | seletz | rereads |
09:33:00 | seletz | maybe i should get some more coffee :) |
09:33:20 | erikm | lol :) |
09:34:11 | seletz | How are you doing on jour new job? |
09:34:42 | erikm | good so far |
09:34:51 | erikm | just doing some userland hacking |
09:35:26 | erikm | quite a difference, but having kernel experience helps a lot |
09:35:45 | erikm | (i.e.: think twice before you code) |
09:36:36 | seletz | well, userland coding is fine too. For me it's just the other way: I did lots of userland coding. But that does not help in while hacking in "kernel mode". |
09:37:20 | seletz | erikm: yup, the "think before code" is needed in serious userland coding as well. |
09:37:26 | erikm | after two years, gtk is still nice, though :) |
09:37:49 | seletz | erikm: eeek! GUI coding :) |
09:38:17 | erikm | seletz: yeah, writing a good GUI is difficult |
09:39:05 | erikm | seletz: but this GUI is for only three technical people, so it can be bad without getting into the interface hall of shame |
09:39:17 | seletz | erikm: well, when i do GUI i do QT, not that i like C++, but i like the way QT works. And the docs are really good. Well, looking at GNOME gtk can't be that bad ... :) |
09:39:31 | seletz | erikm: lol |
09:40:00 | seletz | erikm: it helps when the users are technical educated. Helps a LOT |
09:41:02 | erikm | seletz: I started coding GTK five years ago, when Qt was still a licensing mud. I needed something that had a free license, looked good, and had a better API than Motif |
09:43:25 | seletz | erikm: well, imn my last project in userland we took QT because it's portable and inexpensive. We had to code for AIX, SGI, HP, SUN _and_ NT. |
09:45:04 | seletz | erikm: We have one source tree now, and porting is just an issue because of f***ing weird native C/C++ compilers. Thats fine. |
09:46:34 | erikm | seletz: same for me. we needed something portable. when we started GTK worked on pretty much any POSIX platform, and after two years it even got a native Win32 interface# |
09:47:23 | erikm | seletz: so we got win32 support for free |
09:47:41 | erikm | seletz: not that it really helped us, because win32 doesn't support large files ;) |
09:48:44 | erikm | seletz: the application processed raw video files: 20MB/s, 1.2GB/min |
09:49:05 | seletz | erikm: well, i guess i just did not look hard enough. sigh. Anyway: "never change a running system", and i suppose i won't code some GUIs anytime soon. |
09:49:28 | seletz | erikm: ugh, that's lots of data. |
09:49:41 | seletz | erikm: what did you do? |
09:49:51 | seletz | erikm: the app i mean :) |
09:51:20 | erikm | seletz: image processing. it took two input video streams (taken with two cameras) with a face in it. the face had some blue markers. the app detected markers, matched the markers in both streams, and created a 3D model from it |
09:52:08 | seletz | eeek |
09:52:15 | seletz | runs away in horror |
09:52:35 | erikm | seletz: that was only the easy application :) |
09:52:58 | seletz | erikm: some sort of head tracking for VR stuff i suppose |
09:54:47 | erikm | seletz: the other one took only a single video stream from a face without markers, and had a lot of agents running to process the data. each agent was good in one thing: detecting eyes, mouth, nose, ears, etc. one agent collected all data and created a face description from it. |
09:56:19 | erikm | seletz: the nice thing was that the agents could ask each other things. for example: an agent detecting the position of the left eye wants an estimate for the current eye position. one of the other agents might be able to give that. if not, there was an agent that extrapolated each position from past information |
09:57:55 | seletz | erikm: i never did such weird things. The most complex app i wrote was about creating 3d parts in pro/EINGINEER from an abstract, CAD system independent modeling language. Thos parts could be nested in quite complex ways. The app was user scriptable using perl and highly configurable, had a server mode and so on. But mostly it was simple data processing and coding work around crappy APIs. |
09:58:25 | seletz | erikm: ok, then it was some sort of face recognition? |
09:59:18 | seletz | erikm: agents... sounds cool. Were tose agents processes or "just" a bunch of well defined APIs? |
09:59:25 | seletz | s/tose/those/ |
10:07:46 | seletz | well, SIGLUNCH i'd say ... |
10:21:42 | erikm | sorry, phone |
10:23:18 | erikm | seletz: it was not really face recognition, we only want to get information about the movement of muscles in the face |
10:23:42 | erikm | seletz: the agents were real processes, even running on different hosts |
11:38:06 | seletz | back |
16:30:58 | prpplague | seletz: !!! |
16:31:00 | prpplague | seletz: hows biz? |
16:42:22 | | seletz was last seen on #blob 5 hours, 4 minutes and 16 seconds ago, saying: back [Wed Apr 3 12:38:06 2002] |
16:42:22 | prpplague | ibot: seen seletz |
16:42:39 | prpplague | erikm: well how goes it on your end of the planet? |
16:43:34 | erikm | prpplague: wonderful weather |
16:43:41 | erikm | prpplague: /me heads for a pub |
17:20:34 | seletz | prpplague: woha! |
17:20:41 | seletz | prpplague: are you still here? |
17:21:14 | seletz | damn |
17:23:41 | prpplague | seletz: ya |
17:23:52 | seletz | phew |
17:23:54 | prpplague | seletz: i've been working like mad on a project for the boss |
17:23:57 | seletz | prpplague: hi |
17:24:02 | prpplague | seletz: how goes it? |
17:24:20 | seletz | prpplague: good so far |
17:24:41 | seletz | prpplague: i've been working on a PCMCIA driver for an analogue measurement HW |
17:25:03 | seletz | prpplague: this HW will go into a handheld |
18:13:49 | seletz | see ya |
18:13:55 | seletz | going home |
18:40:35 | seletz | hi again |
18:40:43 | prpplague | howdy |
19:29:07 | seletz_ | got thrown out |