Subject: Heidelberg meeting report. From: Eric Youngdale (eri...@cais.cais.com) Date: Jul 5, 1994 3:48:21 am List: fi.hut.niksula.linux-activists.ibcs I am now home (and a bit jet-lagged) after having gotten back from Heidelberg (Note the time of this post - I have been up for 3 hours now :-). All in all, the meeting was an incredible success - there were 350 seats available and these quickly sold out and they reportedly turned away another 300 people. All of the talks were standing-room only, and there is already talk of another meeting next year in Berlin. This is not really related to iBCS2, but there was a beta version of a Linux-native version of Maple on a machine that people could try out. I have a handout that they were giving out, but it is in German, so I cannot say that much about what they were saying :-). It is not obvious at all in the English-speaking world, but a German computer user benefits from a much greater coverage of linux in the media. This month's issue of iX (a German Unix magazine) has an article written by Drew about the PCI driver, and an article by Bob Amstadt about Wine. I understand that they have been running about one article a month for the past 6 months or so. A second magazine, C't, has also been running about 1 article per issue lately. In addition, there are a number of books available. I walked down the main pedestrian street in Heidelberg and came across a store that sold computer stuff - in the front display window, clearly visible from the street, there were 3 linux books plus the yggdrasil cdrom. One of the books, written by Thomas Uhl and Stephan Strobel, has been translateed into English by Springer, and will soon appear in the U.[S,K]. market. This book is a basic introduction for the new user, so other more advanced books would also be required for a new English-speaking user. The other two German books for linux were more advanced - one contained a bunch of man pages translated into German, and the other one had to do with kernel internals. Pictures were taken of the developers present, and JPEG files should become available soon :-). One thing that became clear fairly quickly was that there is a lot of interest in iBCS2 within the user community. I gather that a lot of people are trying binaries of one sort or another, and a lot of these are never formally reported to us even if the binaries work perfectly. Also, there was a machine set up with both the demo versions of WP and Xess running - this was the first time I had seen either program. Anyway, I now have a few more specific comments. I may have forgotten some, but whatever I forget I can pick up again later. 1) Someone wanted to know about the performance of WP under Linux as compared to the performance of WP under SCO. He was concerned about slowness of scrolling the screen under X, but this could be because it was a slow machine that the demo was running on. Is there someone who has seen this both under Linux and SCO on similar machines who would care to comment? 2) Linus actually asked me about the status of the IPC patches, and he wants to see them at some point. If there are political problems with the previous author of the linux IPC code, I think we can simply go ahead and submit the patches to Linus and let him take the heat. 3) The Xess demo was reportedly dumping core every so often. Have other people observed this? No idea why, but there may be an emulation problem somewhere. If the problem is IPC, it just points the finger in the appropriate direction for us to work. 4) What exactly is the situation with COFF shared libraries. I applied some patches about a month ago from some people - does this basically fix all of the known problems, or are there other issues that remain to be solved? 5) It has been suggested that the iBCS2 emulator is of higher quality than the "ALPHA" label would suggest. Given the large number of people (silently) using it, I would tend to agree. Once IPC is taken care of, I would propose that we go ahead and make it beta. I have already asked Ted T'so for a BETA directory on tsx-11 so that it will be there when we need it. 6) One user of a copy-protected SCO program said that the dongle that came with the program did not work while running the program with iBCS2 under Linux. Not surprising, I suppose, since the dongles tend to do very strange things that could be timing related. He was jokingly suggesting setting up a machine under SCO running as a dongle server. -Eric
Subject: Heidelberg meeting report. From: Eric Youngdale (eri...@cais.cais.com) Date: Jul 5, 1994 5:00:57 am List: fi.hut.niksula.linux-activists.ibcs > Anyway, I now have a few more specific comments. I may have forgotten > some, but whatever I forget I can pick up again later. Ah, yes. As I was looking over my notes, I saw one more relevant point. Linus and I were discussing the problems with major number conflicts (i.e. the sound driver and iBCS2 use the same major number). Apparently some of the QIC drivers also use major==30. Linus does not like static allocations of device majors for things like this, and here is what he proposes: We go to a dynamic major number allocation scheme for things like this where they may not always be used (and the sources are not always with the kernel). The "big" devices, like disks and ttys would use the fixed numbers that are currently used. The idea is that the kernel allocates from some pool and assigns a number. This presents a little bit of a problem WRT the /dev files. He proposes that we create a /proc/dev directory, and the module would create the /proc/dev entries when it is loaded. The /dev/socksys entries would merely be a symlink to point into /proc/dev. The only question I have is WRT ownership and protections of these files, but for iBCS2 this should not be any problem. Ultimately when the devices are "created", we can specify this information, so this affords us some flexibility. Nothing implemented so far - just an idea. How does everyone feel about this? Michael - the changes to your code would be relatively minimal I should think. Once the basic framework is in place, I can show you what is required. -Eric
Subject: Re: Heidelberg meeting report. From: Mike Jagdis (jag...@purplet.demon.co.uk) Date: Jul 5, 1994 4:26:26 pm List: fi.hut.niksula.linux-activists.ibcs > We go to a dynamic major number allocation scheme for things like this > where they may not always be used (and the sources are not always with the > kernel). The "big" devices, like disks and ttys would use the fixed numbers > that are currently used. The idea is that the kernel allocates from some pool > and assigns a number. I did some work on auto assigning device numbers a while back (actually I was looking at various approaches to cloning devices for things like ptys - the auto assignment was incidental). I might still have the patches to fs/devices.c around somewhere. In fact the iBCS registration routines are already designed to assume they will be told who they are if they pass a major number of zero. I seem to remember it working... Mike
Subject: Re: Heidelberg meeting report. From: Mike Jagdis (jag...@purplet.demon.co.uk) Date: Jul 5, 1994 4:42:02 pm List: fi.hut.niksula.linux-activists.ibcs > All in all, the meeting was an incredible success - there were 350 seats > available and these quickly sold out and they reportedly turned away another > 300 people. I feel sort of left out :-(. So left out I dusted off ISODE and PP again and offed 100MB of stuff I didn't immmediately recognise so I could play. I think I might have got the heap corruption nailed this time around :-). > 1) Someone wanted to know about the performance of WP under Linux as > compared to the performance of WP under SCO. He was concerned about slowness > of scrolling the screen under X, but this could be because it was a slow > machine that the demo was running on. Is there someone who has seen this both > under Linux and SCO on similar machines who would care to comment? I have Linux and SCO ODT sitting side by side (well back to back actually) at work. Both 486DX-33 with 16MB RAM and comparable IDE disks. The ODT box has a Cirrus Logic display card, my Linux box has a bog standard ET4000. Both are running WP from an NFS filesystem mounted from another SCO box. Subjectively the Linux box is noticeably faster and more repsonsive than the ODT box. Scrolling WP sized windows is never going to be fast on non accelerated display cards whether Linux or SCO. > 3) The Xess demo was reportedly dumping core every so often. Have > other people observed this? No idea why, but there may be an emulation problem > somewhere. If the problem is IPC, it just points the finger in the appropriate > direction for us to work. I haven't got round to picking up the demo yet. Is it reproducible at all? > 5) It has been suggested that the iBCS2 emulator is of higher quality > than the "ALPHA" label would suggest. Given the large number of people > (silently) using it, I would tend to agree. Once IPC is taken care of, I would > propose that we go ahead and make it beta. I have already asked Ted T'so for a > BETA directory on tsx-11 so that it will be there when we need it. Given the bouncing around witht the last few kernel patches it deserved its "ALPHA" label :-). I think things should have stabilised now though. I think I've said that before too... > 6) One user of a copy-protected SCO program said that the dongle that > came with the program did not work while running the program with iBCS2 under > Linux. Not surprising, I suppose, since the dongles tend to do very strange > things that could be timing related. He was jokingly suggesting setting up a > machine under SCO running as a dongle server. A hardware dongle? Maybe we just need to implement some of the lp ioctls? Not having either the application or the dongle makes this a bit tricky though :-(. Mike
Subject: Re: Heidelberg meeting report. From: johnsonm (john...@merengue.oit.unc.edu) Date: Jul 5, 1994 5:28:18 pm List: fi.hut.niksula.linux-activists.ibcs Eric Youngdale writes: > 3) The Xess demo was reportedly dumping core every so often. Have > other people observed this? No idea why, but there may be an emulation problem > somewhere. If the problem is IPC, it just points the finger in the appropriate > direction for us to work. Hmm. I haven't had it dump core at all, but I'm still using some of the oldest ALPHA patches. I haven't used the new version (the demo version you tried was compiled with recent patches added), which is the first of our SCO line to use Motif 1.2, very much yet, though. I can tell you for sure that the problem is not IPC, since we don't use IPC... If someone can forward any description of anything that caused a core dump, I can try it under SCO and see if I can recreate it there, or if it requires Linux. michaelkjohnson
Subject: Heidelberg meeting report. From: Eric Youngdale (eri...@cais.cais.com) Date: Jul 5, 1994 8:00:13 pm List: fi.hut.niksula.linux-activists.ibcs > I can tell you for sure that the problem is not IPC, since we don't > use IPC... If someone can forward any description of anything that > caused a core dump, I can try it under SCO and see if I can recreate > it there, or if it requires Linux. Oh, crap. I will have to try and remember who it was that told me this. I had assumed that it would be reproducible. Perhaps I will download it and see for myself. -Eric