hedrick@geneva.rutgers.edu (Charles Hedrick) (06/10/89)
This may not be quite the right place for this, but it touches issues that have been discussed here, and I don't think there is yet a group for the OS: FSF folks might want to give some thought to who they expect to use the Gnu OS. The answer to this question will affect some technical decisions and some non-technical policies. From what has been said so far, it seems most likely that the Gnu OS will be Unix-compatible, will be aimed at 32-bit machines, and will not take any extraordinary pains to minimize memory use. (I don't mean it will be intentionally sloppy, just that it won't be like Minix, which has as a specific goal avoid "big" software such as Gnu Emacs and X.) It seems safe to guess that the C library will be covered by the Gnu public license. All of this is fine for me personally. I'd love to use it at home, and to make contributions to the community. None of this would cause me any problems. However there are two potential user communities for which is poses potential problems. First, it cuts out most of our students. By and large they can't afford 32-bit machines, particularly not with enough memory to run "big" Unix software. This may become a non-issue if release of the Gnu OS is delayed enough. At the moment students typically have IBM-compatible systems or Macs. Since FSF has decided to boycott Apple (a decision with which I am sympathetic, by the way), this means that we have to look at IBM compatibles. This is unfortunate, since it means we won't have 32-bit machines until 386's come into the student price range. AT's are just now getting to that point. I think it's still going to be a couple of years for 386's (at least assuming that we're talking about a balanced system, i.e. one with enough memory and disk to make a 386 sensible). Of course if it's two year before the Gnu OS is available, this is a non-issue, but I'd sort of hoped it wouldn't be that long. The second class of users it excludes is installations like Rutgers. I think. Our problem is that there's a good deal of software we want to run that isn't public domain. The Gnu public license is going to prevent commercial software from being ported to the system, because any software linked with the libraries can't be proprietary. This has both its advantages and disadvantages. I see both, and for my personal purposes I actually prefer the FSF position. But one of its costs is going to be preventing the development of a commercial software base. The days when major university facilities could depend entirely upon public domain software are over, though we still try to do so as much as possible. Of course most commerical computer users will have this problem even more. So at the moment it looks like the user base will consist of hackers who can afford reasonably up-scale machines. That includes me, so I'm happy, but I'd sure rather have it include machines that my students and faculty use. One thing that would help a lot would be if the Gnu OS could be made to be binary-compatible with the "native OS" on systems where this is feasible. Mach is apparently already binary compatible with 4.3 BSD on the VAX. Since 4.3 and Ultrix are at least to some extent compatible, this seems to provide access to a reasonable software base on the VAX. It would be nice if something similar could be arranged for SunOS, probably with some limitations (e.g. you almost certainly wouldn't want to support pixrect and the old Sunview, but presumably X-based stuff could be made to work). Presumably for IBM compatibles the natural target is the generic ATT Unix release. If you can afford to wait for System V release 4, you'd have the advantage of something with a degree of Berkeley compatibility. If you could arrange some level of binary compatibility for the major systems, then you'd provide a way to get access to the existing software base, and alleviate some of the problems. I don't have any similarly good suggestion to help student hackers. It probably is too late in the day to write a new OS for the 8086 or 286. One could probably support the Mac by using position-independent code and some software hackery. But of course that would be helping Apple, and we can't do that. Maybe we should start encouraging students to get Amigas? Or maybe FSF could make or at least promote some sort of deal with a small hardware vendor to make available an appropriate 386-based configuration for students. Has FSF considered announcing the machines that are going to be early targets of the Gnu OS. Some people might want to make sure that their next machine is going to be on the list. At the moment about all I know is that machines from Apple aren't going to be on the list.
jeffrey@algor2.UUCP (Jeffrey Kegler) (06/10/89)
In article <Jun.9.17.26.01.1989.1762@geneva.rutgers.edu> hedrick@geneva.rutgers.edu (Charles Hedrick) writes: >From what has been said so far, it seems most likely that the Gnu OS >will be Unix-compatible, will be aimed at 32-bit machines, and will >not take any extraordinary pains to minimize memory use. Myself I hope it is heavily targeted to the 386 market. The 386 with an ISA or EISA bus is as close to an open hardware standard as exists. The user has his choice of vendors, and the ability to mix and match hard disks, monitors, boards, CPU's, mice, etc. The competition in terms of hardware prices is cutthroat. And the reusability of the thing for DOS guarantees a good resale price, for those who have to worry about the quality of the thing as an investment. >First, it cuts out most of our students. By and large they can't >afford 32-bit machines, particularly not with enough memory to run >"big" Unix software. If you price competition to the 386 market I think you will find while base prices may compare, as soon as you talk about any add-ons the prices soar, since there is no competition. A good UNIX box is still beyond a student budget, agreed, but I think best hope lies with the 386. It is already within the budget of a starving consultant like myself. And while there may be good reasons for bringing the GNU OS to a VAX, affordability cannot be one of them. This is not exactly an unbiased opinion. I want to run the GNU OS as soon as it comes out, and I own a 386 with a number of bells and whistles, which runs UNIX. If the first ports of the GNU OS do not include 386's, I am frozen out. More important, however, a 386 port would put the GNU OS into a lot of hands very quickly. -- Jeffrey Kegler, President, Algorists, jeffrey@algor2.UU.NET or uunet!algor2!jeffrey 1762 Wainwright DR, Reston VA 22090