chaiklin@cunixf.cc.columbia.edu (Seth Chaiklin) (06/28/90)
I am runing SCO Xenix/386 2.3.1. I am curious about whether there is a upper limit on uids and also a upper limit on group ids? How do people handle the growing number of uids? Do you fill in the lower values or just keep adding to the end? Does it matter? Thanks, Seth Chaiklin ------------------------------------------------------------------------------ Seth Chaiklin Institute for Learning Technologies (212) 678-3899 Box 8, Teachers College, Columbia University, NYC 10027 INTERNET: chaiklin@cunixf.cc.columbia.edu UUCP: seth@ny-yn ------------------------------------------------------------------------------
syd@DSI.COM (Syd Weinstein) (06/28/90)
chaiklin@cunixf.cc.columbia.edu (Seth Chaiklin) writes: >I am runing SCO Xenix/386 2.3.1. I am curious about whether there >is a upper limit on uids and also a upper limit on group ids? >How do people handle the growing number of uids? Do you fill in >the lower values or just keep adding to the end? Does it matter? SCO, like most Unix's uses 16 bits for the uid and gid. Thus the upper limit is 16 bits. However, when you start to network, you will find some bugs re negative uid's in many Unix's so the practical limit is 32767, not 65535 for the upper limit. Since uid's and gid's should (must if NFS is used) be consistent network wide on you LAN's, perhaps some very large sites might find that constraining. However, for most sites, I would think that approximately 32000 uids and gid's is sufficient. -- ===================================================================== Sydney S. Weinstein, CDP, CCP Elm Coordinator Datacomp Systems, Inc. Voice: (215) 947-9900 syd@DSI.COM or dsinc!syd FAX: (215) 938-0235
nanook@rwing.UUCP (Robert Dinse) (06/30/90)
In article <1990Jun28.031638.15931@cunixf.cc.columbia.edu>, chaiklin@cunixf.cc.columbia.edu (Seth Chaiklin) writes: > > I am runing SCO Xenix/386 2.3.1. I am curious about whether there > is a upper limit on uids and also a upper limit on group ids? > How do people handle the growing number of uids? Do you fill in > the lower values or just keep adding to the end? Does it matter? > Thanks, Seth Chaiklin > I've done some experimenting on that very subject and found 30,000 to be the limit (not 32767 or some sensical boundry). Although you can go this high, to have more about 990 users you have to create seperate /usr or /usr2 or whatever directories, otherwise mkdir blows up. Never ran into a group id problem so don't know what/if max values apply there.
pgd@bbt.se (P.Garbha) (07/02/90)
In article <106@rwing.UUCP> nanook@rwing.UUCP (Robert Dinse) writes: >In article <1990Jun28.031638.15931@cunixf.cc.columbia.edu>, chaiklin@cunixf.cc.columbia.edu (Seth Chaiklin) writes: >> >> I am runing SCO Xenix/386 2.3.1. I am curious about whether there >> is a upper limit on uids and also a upper limit on group ids? > > I've done some experimenting on that very subject and found 30,000 >to be the limit (not 32767 or some sensical boundry). In file /usr/include/sys/param.h you have the line: #define MAXUID 60000 /* max user id */ That would get me to assume the highest uid is 60000, not 30000. The setuid() call makes an explicit test for this, and returns EINVAL if greater. chown does not check anything, and you can have uid's up to 65535
cudcv@warwick.ac.uk (Rob McMahon) (07/03/90)
In article <1990Jun28.165234.23491@DSI.COM> syd@DSI.COM writes: >SCO, like most Unix's uses 16 bits for the uid and gid. Thus the upper limit >is 16 bits. However, when you start to network, you will find some bugs re >negative uid's in many Unix's so the practical limit is 32767, not 65535 for >the upper limit. On the other hand there used to be bugs in assorted utilities with [ug]id's as low as 2-3000. I can't speak for SCO Unix, and we can't get the source these days, so I can't see whether they've all been chased down in the systems that did have these problems. I've seen `lastcomm' dumping core, `ls' taking forever, and `ps' printing numbers instead of usernames, because of static uid->username mapping tables sized at compile-time and indexed by uid. (lastcomm ignored the problem, I think ls kept one entry cached outside 0-2000, and ps just punted.) I'm sure this was true in 4.1BSD and SunOS 2.0. Rob -- UUCP: ...!mcsun!ukc!warwick!cudcv PHONE: +44 203 523037 JANET: cudcv@uk.ac.warwick INET: cudcv@warwick.ac.uk Rob McMahon, Computing Services, Warwick University, Coventry CV4 7AL, England