Date: Thu, 21 Aug 1997 22:44:42 -0700 (PDT) From: Sean Eric Fagan <sef@Kithrup.COM> To: softweyr@xmission.com Cc: chat@freebsd.org Subject: Re: SCO announced SysVr5 Message-ID: <199708220544.WAA05583@kithrup.com>
next in thread | raw e-mail | index | archive | help
>Did they really do anything to warrant calling >this "SVR5", or has the "version number as marketing tool" taken over at >SCO, too? That's a good question. I think a large part for the new version is because it's now a SCO thing. However, their "Open Server" (descendent of "Open Desktop") has diverged significantly since SysVr3.2, and since the new OS is a merge of SysVr4 and it, it kinda makes sense. Also, of course, a big thing is that it will be 64-bit compliant -- including 64-bit file offsets on 32-bit machines. Those are the biggest changes *I* know of. >By the way, Sean, didn't you used to work for the SCO Bozos? I seem to >remember some nasty three-way arguments between you, me, and Terry >Lambert a number of years ago, when Usenet ruled the earth. ;^) Yes, I used to work for SCO, and I would again, if they were doing OS development in Santa Cruz (which they're not, btw). I had problems with some parts of it, but, all in all, I liked the OS. Their 3.2 (aka 3.2.0 and 3.2.1, prior to 3.2v2) was not very good, but that was largely due to them needing (for market reasons) to get a SysVr3.2 product out *NOW*. 3.2v2 was significantly better. I think the SCO OSes were much maligned. I do admit, I really liked Xenix/386 -- it ran X in 4MBytes, for crying out loud! Stable, fast, and small. Wonderful system. (And I did my first "device driver" for that -- /dev/fd, character special nodes that essentially did a dup of the file descriptor corresponding to the minor number.) 3.2v2 was pretty nice, I think. It had working job control (lots of my work there). I still remember one night, I was hacking away at it, trying to get it to work -- and became aware of someone standing behind me. When I looked, I found myself being watched by Doug Michels. (Contrary to popular belief, he did *not* sexually harass me that night. :)) Prior to SCO's acquisition of HCR, in Canadia, I worked on the compiler almost exclusively. I probably knew the Microsoft C Compiler better than all but four or five people in the world. This, incidently, is why I will no longer do compiler internals work :). I still remember some optimization bugs "fondly." :) After the acquisition, I moved to the OS group. The most significant thing I did was adding long filenames and symbolic links to the post-3.2v2 filesytem. That was fun. As practice for it, I also wrote a minimal procfs for the system (it never got released, although I wonder if the code is still on a backup tape somewhere...). The only thing I used it for was to write a renice program -- it did an fcntl() on a /proc file. I know people bitched about lots of things about SCO, but, as I said, I thought it was okay. I know why they did lots of things they did, and why they didn't do lots of things. Their biggest failing as an OS, I think, was that they eventually worked more on features than on tuning the OS. (This, in fact, started with SysVr3.2v3 [which was never released].) So now they've got an OS which causes a 90MHz Pentium with 16MBytes to feel slower than the 25MHz 386 that was the original kithrup. (Of course, that was a development version... hopefully they will spend time tuning it.) And that, of course, gets back to Xenix/286 and Xenix/386. The rate of development was much less than, and Xenix/286 had a period of years in which it was tuned for the PC architecture. So it was *fast*, and *stable*. Pity that can't get done anymore.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199708220544.WAA05583>