From the JAC viewpoint, the key thing (as always) is the ability to check
out the source from CVS and build it on whichever OS we happen to have
chosen post RedHat.

We haven't decided which Linux to adopt so we can still be swayed by
Starlink's choice (since that would make life easier for us). If we decide
on a different vendor then we will just build it locally if that is at all
possible.

The thing that makes the Mac OS X port hard is the fact that Brad is
making lots of little tweaks to all the applications in order to get them
to build properly but has nowhere to send his patches. Next time a package
is updated we will have to make the exact same changes. This is not
sustainable.

Is there yet a "project plan" for putting Starlink Classic into CVS?

The other thing on the horizon is, of course, 64bit linux....

Tim

On Mon, 3 Nov 2003, Giaretta, DL (David) wrote:

> I guess it depends what you mean by support.
>
> The argument in the past (Martin) was that "proper" support is very
> expensive and so we can only do it with lots of effort and several machines
> per OS.
>
> We _have_ to be much more flexible now. Is this at all realistic?
>
> My guess is that this will be sustainable only on the assumption that there
> are no horrible system dependencies in the software - which given that it is
> pretty solid on 2 OSes currently is probably a reasonable assumption.
>
> Given that, then it would lead one to say that we support the software and
> that we expect run-time bugs will be generic ones i.e. will probably happen
> under several/all OSes. Build-time problems will, after configure is sorted
> out, be expected to be few and far between.
>
> All the above is applicable to Starlink classic. For the most part the Java
> code should be reasonably secure - as long as an appropriate JDK/JRE is
> available.
>
> There are several oversimplifications here, but unless the above is broadly
> true then we are in trouble.
>
> Can anyone see any show stoppers?
>
> ...David
>
> -----Original Message-----
> From: Peter W. Draper [mailto:[log in to unmask]]
> Sent: 03 November 2003 13:10
> To: [log in to unmask]
> Subject: Re: The future of RedHat
>
> On Mon, 3 Nov 2003, David Berry wrote:
>
> > > The real question is what could we do if the sites fragment in their use
> of
> > > OSes. Saying that we could not support sites is probably not an option.
> ON
> > > the other hand if we see difficulties with excessive fragmentation then
> we
> > > should point them out to sites.
> >
> > OK, so we wait until a need has been demonstrated for a new OS before
> > jumping into supporting it? And we carry on pressing the benefits of
> > sites having a common operating system (the managers seemed to realise
> > that this was a v. good idea at Preston).
>
> All good sense of course and we should have a recommended flavour of Linux
> (which will probably be Fedora for a while) and be part of what is
> promoted in the UK, but there are more considerations that just what UK
> astronomy sites should use. IRAF supports the following PC versions:
>
>     System                          Distribution    Add'l Systems
>     ------                          ------------    -------------
>     FreeBSD 4.2 and higher          FBSD
>     MacOSX 10.1 and higher          MACX
>     RedHat Linux V6.x thru V8.x     RHUX            Mandrake 7.x and 8.x
>     Slackware V8.x                  LNUX            Debian 2.x, all others
>     Solaris 7 for Intel             SSOL
>     SuSE Linux V6.x thru V7.x       SUSE
>
> that's the long term competition. People don't want to change OS just to
> use our software.
>
> Of course this brings us back to the "./configure;make - empowering users
> to build our software themselves", thread, since I doubt IRAF test for all
> these platforms either.
>
> One other point. Steve has built the USSC for Debian, with little effort
> it seems, that looks good for porting around Linux flavours.
>
> Peter.
>

--
Tim Jenness
JAC software
http://www.jach.hawaii.edu/~timj