RFC: Username mapping for HEAD

[Steve Langasek]

On Thu, 13 Sep 2001, Andrew Bartlett wrote:

> I'm looking into the various ways that usernames can be better mapped in
> the HEAD development branch.  My concern is that the current mapping is
> haphazard and prone to error, and I wish to change it in line with my
> AuthRewrite code.

> I'm thinking about having 3 usernames inside samba:
> username_wire - the username exactly as specified on the wire
> username_mapped - this username after it has been mapped on a global
> basis (username map = )
> username_unix - the username that is actually used on the unix system

> The last of these will be pulled straight from the passdb, which *may*
> in turn be based on the system tables.  This is where forms like
> domain\username and GetPwnam() will be done and where users will be
> dynamically added if nonexistent.  This also allows (given a suitable
> passdb) for us to create more than one user per uid - something I would
> like for build_farm testing.

> Any ideas/comments?

What character set would each username variable be stored in?  I've seen
reports that usernames containing non-ASCII characters[1] currently can't be
used for accessing [homes] shares due to charset problems.  It would be nice
if any reworking of usernames could also address this issue.

Steve Langasek
postmodern programmer


[1] Well, this seems rather a questionable policy on systems where legacy
    software is in use, but Samba shouldn't be one of these "legacy
    applications".