From: Walter Rowe (rowe@cme.nist.gov) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-23 11:08:25 PST In article <13.03.92.130104.210@cogsci.cog.jhu.edu> wjb@cogsci.cog.jhu.edu (Bill Bogstad) writes: >> I can get a number of UNIX products with most of the features of >> BSDI for around $1000. The only thing that BSDI has is source >> code. So you're saying you'd rather buy the OS without the source code than one with it given they are the same price? Why? This is like saying that you'd prefer to take delivery of a car without all the service manuals than one with them. This doesn't make sense to me. If you can get more for the same price, then why not do that. And look at all the tools that the BSDI folks are putting in their release. The attitude a lot of people on this list seem to have is to avoid BSDI. I've seen the product run, and its impressive. Its got most, if not all, the GNU tools, it runs very fast X11R5, the graphics are great, you have support, etc. I can understand the home-hacker being more interested in 386BSD than in BSDI. Its more likely to have problems to solve, which will give the opportunity to hack around and tinker with the kernel. But if you want something which is going to be stable and allow you to get real work done (not that you can't get real work done with 386BSD -- I haven't used it myself), then $1000 seems quite reasonable given all the things that come with BSDI. Walter --- Walter Rowe rowe@cme.nist.gov ...!uunet!cme-durer!rowe #include Message 22 in thread From: Bill Bogstad (wjb@cogsci.cog.jhu.edu) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-23 13:14:25 PST In article rowe@cme.nist.gov (Walter Rowe) writes: >In article <13.03.92.130104.210@cogsci.cog.jhu.edu> >wjb@cogsci.cog.jhu.edu (Bill Bogstad) writes: > >>> I can get a number of UNIX products with most of the features of >>> BSDI for around $1000. The only thing that BSDI has is source >>> code. > >So you're saying you'd rather buy the OS without the source code than >one with it given they are the same price? Why? This is like saying >that you'd prefer to take delivery of a car without all the service >manuals than one with them. Personally? Hell no. I'ld much prefer an OS with sources for my own use. Professionally, there are other considerations which may be more important then having sources. Like support for the "standard" 386 UNIX ABI and/or "important" (means what the enduser wants) commercial software packages. >This doesn't make sense to me. If you can get more for the same >price, then why not do that. And look at all the tools that the BSDI >folks are putting in their release. The attitude a lot of people on >this list seem to have is to avoid BSDI. I've seen the product run, >and its impressive. Its got most, if not all, the GNU tools, it runs >very fast X11R5, the graphics are great, you have support, etc. I've only seen it demoed at USENIX and I'm still impressed. You get more of one thing and less of another. And don't forget your mileage may vary... >I can understand the home-hacker being more interested in 386BSD than >in BSDI. Its more likely to have problems to solve, which will give >the opportunity to hack around and tinker with the kernel. But if you >want something which is going to be stable and allow you to get real >work done (not that you can't get real work done with 386BSD -- I >haven't used it myself), then $1000 seems quite reasonable given all >the things that come with BSDI. I can see the home-hacker being more interested in 386BSD. I can see the pseudo-home-hacker being more interested in BSDI (less work). I can see a businessman who just wants a supported system to run commercial applications giving both of them a pass and going with any of the various System V R4 vendors. Bill Bogstad By the way, it was suggested to me in private email that BSDI plans to support the standard 386 ABI in their first non-Beta release sometime this summer. Would anybody care to confirm or deny this rumor? If true this would (IMHO) make BSDI much more interesting to business users. Message 23 in thread From: Sean Eric Fagan (sef@kithrup.COM) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-23 14:20:41 PST In article <23.03.92.161425.216@cogsci.cog.jhu.edu> wjb@cogsci.cog.jhu.edu (Bill Bogstad) writes: >By the way, it was suggested to me in private email that BSDI plans to >support the standard 386 ABI in their first non-Beta release sometime this >summer. Would anybody care to confirm or deny this rumor? If true this >would (IMHO) make BSDI much more interesting to business users. At UseNIX, Rob Kolstad promised to be SCO compatible. In private discussions with him, I verified that BSDinc plans on making their product both DOS (which version? I dunno) and iBCS2 compatible. (iBCS2 being the version of the BCS Intel, SCO, and AT&T hammered out which contains SCO compatibility.) BSDinc has a very, very, *very* bright bunch of folks working for them, and if anybody can do it, they can. My doubts about it are because I am a pessimist, and don't think it'll be anywhere near as easy as they think it is (and they think it's not-easy 8-)). The hardest part of that is, to me, making sure all of the ioctl's work properly. (SCO has a lot of ioctl controls for the video and keyboard, as well as some stuff for generic tty input/output.) -- Sean Eric Fagan | "One form to rule them all, one form to find them, one sef@kithrup.COM | form to bring them all and in the darkness rewrite the -----------------+ hell out of them" -- sendmail ruleset 3 comment from DEC. Any opinions expressed are my own, and generally unpopular with others. Message 24 in thread From: Bill Bogstad (wjb@cogsci.cog.jhu.edu) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-23 15:31:53 PST In article <1992Mar23.222041.7264@kithrup.COM> sef@kithrup.COM (Sean Eric Fagan) writes: >In article <23.03.92.161425.216@cogsci.cog.jhu.edu> wjb@cogsci.cog.jhu.edu (Bill Bogstad) writes: >>[Is BSDI supporting binary compatibility?] > >At UseNIX, Rob Kolstad promised to be SCO compatible. I was there and had interpreted the statements made as they were planning to work on it, but no statements about delivery... Sean's further comments and the files on ftp.uu.net pointed out to me in another posting confirm that they are planning to have this support in the first non-Beta release AND the release is expected this summer. I wish somebody had called me on this earlier as my biggest concern about BSDI was this issue. It's really starting to look like they will have an extremely competitive commercial package. It will still be a little expensive for most hackers, but the added bonus of full sources may make up even for that. Bill Bogstad wjb@cogsci.cog.jhu.edu Message 25 in thread From: Tim Russell (trussell@isis.cs.du.edu) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-23 17:34:28 PST wjb@cogsci.cog.jhu.edu (Bill Bogstad) writes: >In article <1992Mar23.222041.7264@kithrup.COM> sef@kithrup.COM (Sean Eric Fagan) writes: >>At UseNIX, Rob Kolstad promised to be SCO compatible. >I was there and had interpreted the statements made as they were >planning to work on it, but no statements about delivery... Well, the writeup in the Mar/Apr issue of SCO Magazine says "Support for SCO applications is planned by fall." Doesn't sound to me like the first production version will have it. -- Tim Russell Omaha, NE trussell@isis.cs.du.edu Message 26 in thread From: Paul A Vixie (vixie@pa.dec.com) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-23 17:32:48 PST > The hardest part of that is, to me, making sure all of the ioctl's work > properly. (SCO has a lot of ioctl controls for the video and keyboard, as > well as some stuff for generic tty input/output.) I don't agree. The hardest part is all those binaries that were compiled against brain-dead include files that specify 14-character maximum file names. After that the hardest part is all the compiled-in pathnames to things that have moved in 4.4bsd and for which no sane person wants to put symlinks -- there aren't enough airsickness bags on this planet to make it all fit. Ick. On the other hand, if anybody can make 90% of the commercial 386ABI-ware run on a 4.4bsd-based 386 kernel, BSDI is the one. -- Paul Vixie, DEC Network Systems Lab Palo Alto, California, USA "Ready, Fire, Aim" decwrl!vixie vixie!paul Message 27 in thread From: Sean Eric Fagan (sef@kithrup.COM) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-23 18:59:43 PST In article vixie@pa.dec.com (Paul A Vixie) writes: >I don't agree. The hardest part is all those binaries that were compiled >against brain-dead include files that specify 14-character maximum file names. That's easy. Write a v7-ish filesystem. I've heard (but not verified) that somebody has written a XENIX filesystem for bsd. >After that the hardest part is all the compiled-in pathnames to things that >have moved in 4.4bsd and for which no sane person wants to put symlinks -- >there aren't enough airsickness bags on this planet to make it all fit. That also isn't that difficult, as most of them are the same for both BSD and SysV bases. But there are also things which don't exist in BSD, and will need to be rewritten. Terminfo, for example. Not an insurmountable problem, true, but not necessarily trivial. SCO's C2 .... stuff. The networking. STREAMS (vs. streams and bstreams). MultiScreens(tm), and the various controls that go with it. The international support SCO added, both to the kernel and to applications (very interesting stuff, that, actually...). All of these need to be done for applications, as well as a whole slew of other things that I can't remember or think of right now. When I brought up the subject of compatibility at UseNIX, Kolstad implied they would be 100% compatible with SCO. This is rather difficult, as some of the things aren't necessarily documented. >On the other hand, if anybody can make 90% of the commercial 386ABI-ware >run on a 4.4bsd-based 386 kernel, BSDI is the one. Please... it's a BCS, not an ABI. There is a big difference between the two. -- Sean Eric Fagan | "One form to rule them all, one form to find them, one sef@kithrup.COM | form to bring them all and in the darkness rewrite the -----------------+ hell out of them" -- sendmail ruleset 3 comment from DEC. Any opinions expressed are my own, and generally unpopular with others. Message 28 in thread From: Ted Lemon (mellon@hansen.ncd.com) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-24 13:48:52 PST >That's easy. Write a v7-ish filesystem. I've heard (but not verified) that >somebody has written a XENIX filesystem for bsd. I think I'll be needing that airsickness bag now... >That also isn't that difficult, as most of them are the same for both BSD >and SysV bases. But there are also things which don't exist in BSD, and >will need to be rewritten. Terminfo, for example. Not an insurmountable >problem, true, but not necessarily trivial. In BSD, they've split /etc into /etc and /sbin, where /etc contain only data files, and /sbin only contains binaries; likewise for /usr/etc and /usr/sbin. You probably don't want to symlink everything in /sbin into /etc just to satisfy some bogus application that has hardwired the location of some file into its binaries. Fortunately, that's probably not a big problem until you start talking about really big, ugly applications like RDBMSs, which tend to do a lot of thinking about what's where. In most cases, hacking the install script will get around this. If you've seen any such install scripts, you're probably chomping at the bit for the chance to do that - it's *such* fun. On the other hand, there is also the problem of sifting through the kernel trying to find out about how a system is configured. Getting around this problem is probably a major challenge, and particularly depressing in that most such vendor-provided programs run only once - at install time.  _MelloN_ Message 29 in thread From: Peter da Silva (peter@ferranti.com) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-25 15:30:55 PST In article mellon@hansen.ncd.com (Ted Lemon) writes: > In BSD, they've split /etc into /etc and /sbin, where /etc contain > only data files, and /sbin only contains binaries; likewise for > /usr/etc and /usr/sbin. Of course, if they really wanted to make things convenient they'd do it like so: /etc/binexecutables(login, getty, init, cron...) /etc/libfixed data files(termcap, perms, ...) /etcconfig and logging(passwd, inittab, utmp) And same for /usr... -- -- Peter da Silva, Ferranti International Controls Corporation -- Sugar Land, TX 77487-5012; +1 713 274 5180 -- "Have you hugged your wolf today?" Message 30 in thread From: Michael J Hopkirk (hops@x.co.uk) Subject: Re: 386BSD announcement Newsgroups: comp.unix.bsd View this article only Date: 1992-03-27 06:12:59 PST >>>>> On 25 Mar 92 23:30:55 GMT, peter@ferranti.com (Peter da Silva) said: PdS> In article mellon@hansen.ncd.com (Ted Lemon) writes: > In BSD, they've split /etc into /etc and /sbin, where /etc contain > only data files, and /sbin only contains binaries; likewise for > /usr/etc and /usr/sbin. PdS> Of course, if they really wanted to make things convenient they'd do PdS> it like so: PdS> /etc/binexecutables(login, getty, init, cron...) PdS> /etc/libfixed data files(termcap, perms, ...) PdS> /etcconfig and logging(passwd, inittab, utmp) Yet another fs layout - how wonderful.... The names are similar to SysVr4. If the layout is supposed to be the same /sbin - essential executables for admin and ops /etc - machine specific admin config files and system admin databases  ( non varying ) /var - filesystem for varying files ( log files )  ( system and user ) PdS> And same for /usr... /usr - architecture dependent files /usr/bin - majority of system utiities /usr/lib - program libs and architecture dependant databases /usr/share - architecture independant sharable files ( glitch ) PdS> -- PdS> -- Peter da Silva, Ferranti International Controls Corporation PdS> -- Sugar Land, TX 77487-5012; +1 713 274 5180 PdS> -- "Have you hugged your wolf today?" -- Everything disclaimed (including disclaimer) ------< hops@x.co.uk >------< hops@ixi.uucp >------< ...!uunet!ixi!hops >---- Mike Hopkirk ( hops ) | Motif and X - kludge piled on kludge IXI Ltd ! There MUST be a better way...