hsotname stuff

[Robert Olson]

At 02:33 PM 3/26/2004, Ivan R. Judson wrote:

> > -----Original Message-----
> > From: owner-ag-dev at mcs.anl.gov
> > [mailto:owner-ag-dev at mcs.anl.gov] On Behalf Of Robert Olson
> > Sent: Friday, March 26, 2004 1:38 PM
> > To: ag-dev at mcs.anl.gov
> > Subject: hsotname stuff
> >
> > The hands-off mechanism for local hostname determination is
> > now in place, on windows anyway.
> >
> > If a GLOBUS_HOSTNAME is set, its value is left in place
> > without checking (bug 888).
> >
> > This  brings up another issue: it appears that throughout,
> > SystemConfig.Gethostname is used; that uses socket.getfqdn()
> > which does not benefit from the accurate hostname
> > determination. We should either use the globus gethostname,
> > or we should make systmeconfig use the globus gethostname.
>
>I'd suggest that we leave systemconfig.gethostname doing what it's doing,
>but document the heck out of the fact that it's only for verification, that
>we should be using globusconfig.gethostname throughout the toolkit instead.
>
>Does that solve the problem? The reason is, I'd like to know when the system
>hostname is not the globus hostname, in some cases.
>

Yes, if we change the instances of systemconfig.gethostname to 
globusconfig.gethostname.

> > Another related issue is this:
> >
> >          s = ThreadingGSISOAPServer(addr, config = self.config)
> >
> > that is, the creation of the servers passing in the current
> > hostname. I suspect that if the constructor is working
> > properly, we will find that such a server would only listen
> > on one interface of a multihomed computer.
>
>That is the correct behavior, or there's a modification that we need to make
>to make it correct?

I don't think that would be the correct behavior; if I was running such a 
service, unless I wanted it to I'd assume it was listening on all 
interfaces. I think the right thing is to not pass a hostname in, which 
would make the listener take the default behavior of listening on all 
interfaces.

--bob