Norman,

>Up to a point.  That's true as long as dependencies on specific
>versions can be adequately encoded in the RPMs.  Otherwise, someone can
>download some spanking new tarball of a component and then complain to
>us and the world that It Doesn't Work because they were able to do that
>without downloading and installing all the latest versions of the
>components it depends on.

But that applies to most software, most software packages have
dependencies, and not all are included my default on a system.

Steve.


-----Original Message-----
From: Starlink development [mailto:[log in to unmask]] On Behalf Of
Norman Gray
Sent: 25 April 2005 10:54
To: [log in to unmask]
Subject: Re: Advertised location for source distributions

Tim and Steve (and all),

On 2005 Apr 22 , at 22.53, Tim Jenness wrote:

>> Nightly build tarballs are good PR, and are important for
>
> AST will probably be a special case since people want the newest
> features
> for testing

That still doesn't imply nightly builds.  It probably is encouragement
for DavidB to make frequent -- possibly very frequent -- minor
releases, but I don't think it implies that every night's build should
be regarded as one for the waiting millions.

I agree with Steve that since the nightly builds are ready to hand,
it's both good PR and no hassle to make them publicly available.
However nightly build tarballs are not releases.  As far as I'm
concerned -- and my point is that I'm sure I'm not the only one -- the
_only_ Starlink software available on the web pages is the Summer 2002
release, and the nightly builds simply don't count.

Steve:
> The concept of a release is a bit fuzzier now. Now we can update
> libraries and the release with individual package RPMS etc on a
> continual basis. We are not restricted by having static libraries.
>
> The only thing you developers have to keep in mind is that you have to
> tell me when you would like to release a new version of a package,
only
> you lot know that.

Up to a point.  That's true as long as dependencies on specific
versions can be adequately encoded in the RPMs.  Otherwise, someone can
download some spanking new tarball of a component and then complain to
us and the world that It Doesn't Work because they were able to do that
without downloading and installing all the latest versions of the
components it depends on.

AST is currently at 3.6.  Perhaps that minor version number (for
example) should climb rather rapidly, and indeed climb whenever David
thinks there's a version that other folk can reasonably be given to
play with.

CVS 'tag' actions can, if I recall correctly, have hooks.  It might be
easy to attach a `create tarball' action to any time that a component
(such as AST) was tagged with a 'cpt-major-minor' tag name.

Don't misunderstand me: I think it's good that the nightly builds are
available.  I just think we shouldn't relax, since as long as
perfectly-functioning nightly builds are all that are available, then
as far as the outside world goes, we haven't produced anything new at
all.



Steve:
> It is also the "normal" way of distributing source for
> open source projects, plus you do not have to bootstrap them,
> bootstrapping them will probably be unfamiliar to most people.

Tim:
> but they want to build without using the Starlink starconf.

I don't understand.  As ever, the only way that starconf has ever got
involved is when you're building a checkout.  It has never, ever, been
the case that anyone doing the './configure;make;make install' ritual
has ever had to know anything about starconf at all.  Why are we
talking about bootstrap and starconf?  Are we talking at horrendous
cross-purposes?

Steve:
> Yes the "stable" release source and built versions will be advertised
> on
> the main Starlink site as normal. Do you consider everything in CVS
OK?

Me?  Yes, as far as I'm concerned these have been OK for ages, apart
from the bugs which the nightly builds have exposed.  That is, I for
one have no plans to add features, and see no major omissions.

> 's definitely time for a real bona fide new Starlink release.

Oh, yes.  Yes indeedy.

See you,

Norman


--
----------------------------------------------------------------------
Norman Gray  :  Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/  :  www.starlink.ac.uk


--
----------------------------------------------------------------------
Norman Gray  :  Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/  :  www.starlink.ac.uk