Al,
> > > Basically we're going to move to CVS, the next release isn't until the
> > > autumn. If we don't have the classic software in CVS before then we
> > > may as well all give up and go home. The next release will be CVS
> > > based, why put effort into something that won't get used.
> >
> > Because we need to be able to create and distribute usable systems in
> > between full releases. What if someone (e.g. JACH) requests urgent new
> > features - do we say "sorry you can't have them until the next official
> > full release - and maybe not even then if problems crop up with the new
> > system (which you have got to admit is not unlikely)"?
>
> Pretty much, yes (sorry Tim)... if they're going to cut our funding to
> this level we're going to have to cut some corners, this is a corner that
> needs cutting. If we all concentrate on it we can get everything into CVS
> inside a month.
I must confess to being a *little* sceptical... Given the time I have
already put into sorting AST out, given that Norman had previously got it
almost there before I even began, and given that we are still some way of
being able to send a new-style ast.tar.gz to anyone.
> > I do not think we should pull the carpet out from underneath us until we
> > have a working, fully implemented carpet to *immediately* put in its
> > place.
>
> I think we have to concentrate all our efforts to create the new carpet,
> and forget about keeping the old one clean.
But what if *major* problems crop up with the new system, such as
for instance, it taking a factor 5-10 more effort to get going than
previously thought? And even when we have a basically working CVS
system, do you honestly believe that it will be at a sufficient level of
maturity and robustness to be able to safely totally junk the old system
for some time?
> Policy decision I think...
As a policy decision, throwing away our one and only working product in
the *hope* that we will have an equally robust system available to replace
it by the time of the next release is a bit of a high risk policy! The new
system is (to us) new and still in a state of high flux, and will be for
along time. Until it has prooved itself in the field rather than just in
the laboratory, I think it would be "shear folly" to throw the old system
away. We need a backup.
And if we end up unable to deliver a reliable working system of any kind,
then it wont be just corners that are cut...
Anyway, what actual disadvatages are there to keeping /classic going
as it is at the moment? What do we gain by throwing it away?
David
|