Re: Mapping of RIDs to uid_t and gid_t

Luke Kenneth Casson Leighton (lkcl@switchboard.net)
Sat, 4 Apr 1998 17:00:32 +0100 (BST)

Date:	Sat, 4 Apr 1998 17:00:32 +0100 (BST)
From:	Luke Kenneth Casson Leighton <lkcl@switchboard.net>
To:	Multiple recipients of list <samba-technical@samba.anu.edu.au>, Jeremy Allison <jallison@whistle.com>
Subject: Re: Mapping of RIDs to uid_t and gid_t
In-Reply-To: <3525355E.237C228A@whistle.com>

On Sat, 4 Apr 1998, Jeremy Allison wrote:

> Allan Bjorklund wrote:
> >
> > This will not work at our site. We currently have
> > 85060 active entries in our global password file
> > (and growing).
> >
> > We have to use 32 bit UIDs and GIDs.
>
> Well I wasn't excluding sites that are using
> 32 bit uids and gids, just ones that are using
> them greater than 2^28 - 1.
>
> Last time I looked 85060 is less than 2^28-1 :-)
> so you should still be able to use the scheme, so
> long as the uids and gids you have allocated
> don't use the top 4 bits.
>
> I don't have much of a choice here - I have
> to give a 32 bit token back to NT and that's
> all I will get back from it in order to unambiguously
> differentiate a user or group.

ok... so... you get a unique RID back (unique across the entire space of
RIDS, which includes groups _and_ users). why not hold the entire table
of RIDS->unix uid/gids in shared memory, if there is an efficiency issue
for reverse-lookups of NT RIDs to unix uid/gids.

luke