On Thu, 2007-05-03 at 22:34 +0100, Peter Tribble wrote:
> > This is a difficult question...
> >
> > On Wed, 2007-05-02 at 22:12 +0200, Thomas Wagner wrote:
> > > as an example, if I want to create a SMF entry for a daemon.
> > >
> > > To run the daemon with a specific, non-provileged userid, a
> > > separate user has to be created and the name will be stored
> > > in a SMF-propery.
> > >
> > > How should a SFE package behave at pkgadd:
> > >
> > >  1) if the userid is missing, just create it with useradd
> > >     and let solaris choose the next free numeric ID
> >
> > I'm sure that many sysadmins won't like this.  Like you say
> > below, they can create it before installing the package,
> > but how do they find out that they are supposed to do that?
> 
> What I don't like is having to do something outside the
> package installation process. Packages already choose
> usernames and userids, so we're already used to it.

Fair enough.  This may be the best solution then, after all.

> > >  2) ask the user about the username / userid
> >
> > No, interactive package installation is not a good idea.
> 
> Well, for standalone installation, I usually run it interactively
> the first time anyway. And then I could use pkgask to
> create a response file. So I'm quite happy with this
> approach.

I'd like to keep SFE packages "Solaris compatible" and
interactive package installation is a big no-no in Solaris,
so I'd rather not use pkgask.

> > > I would like to do 1) as many (all?) distros do. If an
> > > Administrator does not want the generated userid, he can
> > > create it before pkgadd.
> >
> > Perhaps another option is a preinstall script that checks
> > if the user id exists and if not, if stops with a message
> > explaining what to do and exit 1.
> 
> I don't like that - I would much rather have it just create it or
> let me supply a response file.

Understood.  My vote is 1) then.

Thanks,
Laca


Reply via email to