The original question was what problems can I expect as I exceed 2^15
UIDs (the uid_t in many systems is short). Thanks go to:
"Karl E. Vogel" <vogelke@c17mis.region2.wpafb.af.mil>
Dan Pritts <danno@ans.net>
"Mike Wilson" <mikew@onramp.net>
Jochen Bern <bern@TI.Uni-Trier.DE>
Casper Dik <casper@holland.Sun.COM>
Bryan Hodgson <bryanh@astea.com>
David Thorburn-Gundlach <david@bae.uga.edu>
One person suggested looking in the archives, which I should have done
before posting (duh!). Searching turned up a summary with infodoc 14617
which states various problems with large UIDs. If you want the details,
I suggest searching the archive. :)
I did get some interesting info that wasn't in the archive, though,
most notably that Solaris didn't support large UIDs until 2.5.1,
AIX has always supported it, and HP is supposed to have a fix coming.
Interestingly, the infodoc mentions that "SunOS 4.x systems running the
NFS version 2 software..." making it unclear whether NFS v2 also
has a problem with long UIDs, so I looked up the RFC (1094) and
it defines uid as unsigned int, so this may be a problem on
systems with 2-byte ints (ugh!) although it's probably really a
4-byte field and is implemented as long on 2-byte int systems.
Even if I stuck with all Solaris >=2.5.1, or all systems with a long
uid_t, I would still run into problems with quotas and tar/cpio/ar as they
have their own uid limits (described in the infodoc).
Two people suggested using duplicate UIDs and making sure no duplicates
were on the same filesystem, to avoid quota problems, but then I would
probably have problems with mail, and definitely security issues, so
that is not possible.
-- ~frank * Yoda of Borg I am. Futile resistance is. Assimilated you will be. * * PGP ID: C001AA75 -|- fcusack@voicenet.com *
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:30 CDT