More utmp stuff

Elrond Elrond at
Mon Mar 27 16:34:01 GMT 2000

On Mon, Mar 27, 2000 at 04:23:33PM +0100, David Lee wrote:
> > You should realy consider taking a deep look at screen and
> > rxvt (in that order). There's no need to reinvent the wheel
> > every time, and they did much of that work already.
> > (and they're GPL'ed, so there shouldn't be any
> > copyright-issues, or do I miss something?)
> Thanks.  Do you have any pointers, URLs etc.?

You might also take a look at for some
ssh-implementations, but since ssh (especialy the
"original" has some copyright-issues, you should only use
this as a source of information, not to take code from.)

Could someone from the samba-team please confirm, that
taking parts from screen and including it in samba will not
violate any copyright-issues in samba or related to samba?
So David Lee and who ever helps him can start out taking
source from there (as they see fit of course)

> > > o  Despite the above downgrading of "utmp dir" to geek status, there
> > >    should probably be a "wtmp dir" too.
> > [...]
> > 
> > I don't know, if there's any system out there, that uses
> > something else then a single file for utmp/wtmp, so maybe
> > the options should be "utmp file" and "wtmp file".
> > (wtmp file = /var/log/wtmp.samba)
> The reason I went for "utmp dir" was that it would cover all four possible
> files across many UNIX systems: utmp, utmpx, wtmp, wtmpx .
> The basic idea was to be as simple as possible for "plug and play" use. 
> Ideally, simply "utmp=yes" would be needed, using the OS's own default
> mechanisms and locations.  "utmp directory" was to allow relocation of
> path, but still use the default (relative) filenames and mechanisms. 
> Since then, the suggestion has arisen also to have a "wtmp dir" 
> possibility.
> Ideally, Samba would work out sensible defaults (as, apparently, do your
> suggested "screen" and "rxvt" packages), so that "utmp dir" and any "wtmp
> dir" would not be needed at all by most sys.admins.

That's of course the major idea.

My idea behind this was that some sysadmins like to have
different *tmp-files for different things, and some of them
like to put them in the same directory.

> One of the problems about specifying the final component filename (in your
> example the "wtmp.samba" component) is that several systems maintain both
> the non-x- and x- files in parallel, assuming that they can simple add or
> delete a final "x".  Thus, I think a "wtmp.sambax" (note file "x") would
> magically, but wrongly, appear, wouldn't it?

I haven't got an OS handy which has x-files, so I can't
play with that. But I wouldn't mind it, if there would be
a wtmp.samba and a wtmp.sambax.

> > > o  Worthy consideration should be given to the idea of consolidating
> > >    multiple connections from the same PC into one single utmp record,
> > >    perhaps by a "utmp consolidate" parameter, and idea suggested and
> > >    already implemented by "Freddie <freddie at>" .
> > [...]
> > 
> > But it should only merge entries for the _same_ user.
> This isn't an area I know much about.  When a PC makes multiple
> connections to a Samba server, isn't it forced to use the same UNIX
> identifier (i.e. user) for all connections anyway? 

I don't know much about multiple users on the same
tcp-connection (but this is AFAIK possible!). But:
You of course can have different users on different
tcp-connections, and the PC could run some terminal-server,
that is configured to make distinct connection for each
session, or it could even be another Unix-box with smbfs or


More information about the samba-technical mailing list