[ILUG] numeric user names

[Brian Foster]

  | Date: Thu, 19 Oct 2006 12:33:31 +0100
  | From: jm at jmason.org (Justin Mason)
  | 
  | Pádraig Brady writes:
  | > Does anyone know of existing systems that use numeric usernames?
  | > 
  | > There is a proposal to change `chown 1234 file` in coreutils
  | > to not look for a user name 1234 first, and just assume one means
  | > the user id 1234.  [ ... ]
  | 
  | I can't think of one -- but the principle of least surprise would
  | dictate at least *checking* getpwnam() first.

 back when I was working on sys admin and security tools
 for a Unix company a long long time ago in a place far
 far away, this issue came up.  the same idea (back then)
 was driven by getpwnam(3) being slow.

 I don't recall the arguments used against the idea then,
 and the only rationale I can recall being used for the
 idea was that getpwnam(3) was slow (esp., AFAICR, if the
 NIS/YP/whatever server was inaccessible).  hence, if my
 memory is correct, the alternative which I _think_ was
 ultimately implemented (cannot recall for certain now):
  an option (`-n' or something, cannot recall now):  if
 `-n' was given, then the "name" had to be numeric (an
 error otherwise), and was taken as the numeric UID.

 also --- what I can recall more slightly clearly (since
 I did or reviewed some of this work meself) --- is the
 account creation tools (equivalent of useradd(8) &tc)
 refused to create a numeric account.  the tool which
 checked the various files for consistency (`pwcheck' or
 something, not sure if there is a Linux/GNU equivalent?)
 whinged if it found a numeric username (or groupname).

 however, the tools that modified an existing account
 (e.g., passwd(1) &tc) were quite happy to work with
 numeric usernames.  hence, you had to "roll yer own"
 to create such beasties, but if you had created them,
 you were told this is a bad idea yet all the nominal
 stuff still worked.

 my memory is rather flakey here, and I'm not claiming
 the above is 100% accurate, but that's how I recall the
 same idea(/issue?) a universe away being dealt with.
 this was all pre-POSIX, using K&R-1 C, which gives you
 an idea just how long ago this was ....  ;-\ 
cheers!
	-blf-
-- 
Experienced (20+ yrs) kernel/software Eng: | Brian Foster   Montpellier,
 • Unix, embedded, &tc;  • Linux;  • doc;  | blf at utvinternet.ie   FRANCE
 • IDL, automated testing, process, &tc.   |  Stop E$$o (ExxonMobile)!
Résumé (CV) http://www.blf.utvinternet.ie  |     http://www.stopesso.com