Patches for Solaris 8 compile

[David Lee]

On Wed, 7 Mar 2001, Jeremy Allison wrote:

> Toomas Soome wrote:
> > 
> > ah, at last some feedback:)
> >
> > the alternative way is to let utmp logging as it is and use the same idea
> > as dtlogin (CDE login) does in solaris - dtlogin does register sessions on
> > line /dev/dtremote, so you can see logins on line dtremote. and:
> > 
> > [144] tsoome at madli:hs/txt> ls -l /dev/dtremote
> > lrwxrwxrwx   1 root     root           9 24. apr  2000 /dev/dtremote ->
> > /dev/null
> > 
> > so, I found this idea to be able to see smb sessions with finger to bee
> > good and since it only will require few lines of code to implement, why
> > not?
> 
> Ok - I spoke to Andrew about this and he has the same
> reaction that I did - we *really* don't want Samba creating
> stuff in /dev - not even if it's a symlink to /dev/null.

Firstly my huge apologies, especially to Toomas Soome, for not replying
earlier on this thread, which I recall he started before Christmas. 
Sorry. 

I write simply as the person who cobbled together Samba's umtp support,
not with any great knowledge of Samba, nor to defend my less-than-perfect
code. 

I seem to recall that the "finger" problem was spotted last summer, but
that the consensus was to live with it as a curious feature.

> It's one of those "we have no business doing things there"  issues that
> will cause problems with administrators. 

As I understand (I may be wrong), "utmp*" support requires some sort of
"line".  This is what the "/dev/smb/1" invention is: in one sense, simply
a marker.  But I also tend to agree with Jeremy et al that Samba probably
should not be diddling with such (pseudo-)devices in the system.

One vital thing to bear in mind, is that utmp* support is incredibly
varied across UNIX systems.  The portability considerations are large, as
I discovered the hard way!

Note that I invented "/dev/smb/1" as a subdirectory.  This follows the
Solaris model (but note that other systems have a different model: see the
source code).  I suppose that, being a self-contained directory, it would
at least constrain any activity, were the decision made to create such
entities.

Perhaps there could be yet another parameter: boolean "utmp create line",
propably defaulting off, to allow brave sys.admins to configure their
machines to create these entities.  Might that help? 

Hope that helps.

-- 

:  David Lee                                I.T. Service          :
:  Systems Programmer                       Computer Centre       :
:                                           University of Durham  :
:  http://www.dur.ac.uk/t.d.lee/            South Road            :
:                                           Durham                :
:  Phone: +44 191 374 2882                  U.K.                  :