user ID

[Peter Silva]

In article <1991May8.202912.25142 at bnlux1.bnl.gov>, bstewart at bnlux1.bnl.gov
(Bruce Stewart) writes:
|> 
|>   The System Administrator's Guide states that a user ID number must
|> be between 0 and 60,000; is this for real? Most systems  have a
|> maximum of 2^16-1 = 65535 which is understandable. If 60,000
|> is for real it will cause some serious grief to our computing
|> services division which is trying to establish uniform user
|> IDs for all our systems.

Wait! There's more!

	EP/IX  (like MIPS/OS, but value added by CDC) says
all uids are < 50000.

And EP/IX, IRIX, and SunOS all disagree on who nobody is
   
uid and gid = 14, 30001, and 65534 respectively (this doesn't
matter unless you use NFS, which you probably do! )  On 
SunOS, 65534 used to be -2, but POSIX says uids are unsigned, so SUNOS
changed the interpretation of the same bit pattern.
We have a mix of SunOS 4.0.x and 4.1.x (not to mention
PC-NFS) running, and we are constantly bombarded by messages
from the accounting system complaining about this.

I prefer 14 myself, but try convincing other OS's about this
wisdom...

Any other pearls of inter-operability out there ?


--
Peter Silva			OS Support 
psilva at cid.aes.doe.ca		Dorval Computing Centre
(514) 421-4692			Atmospheric Environment Service