Re: Modest proposal for successful releases

["Richard G. Roberto" <richr@Bear.COM>]

On Sat, 18 Jan 1997, Christoph Lameter wrote:

> On Sat, 18 Jan 1997, David Engel wrote:
> 
> david >On Fri, 17 Jan 1997, Dan Stromberg wrote:
> david >> I can't believe I'm saying this, but I think this means too much of the
> david >> year is spent in testing.
> david >> ...
> david >
> david >I hate to be a "me too'er", but I agree that dropping back to 2 or 3
> david >releases a year would be a good thing.
> 
> The problem of releases is more in the coordination of things than in
> "testing". Basically we rework upstream stuff which is already tested.
> Most problems we have is coordinating changes and making sure it all fits
> together.

I think you are taking too lightly the task of integrating all of
this.  In many ways integration is a more demanding and complex
than software engineering, although useless without it.

> 
> Perhaps we could have a small group of people who are responsible for the
> release. They could make changes on their own to the distribution and/or 
> set deadlines for developers to fix critical bugs.
> A dependency that is off or other minor issues need to be fixed
> immediately rather than wait for the maintainer to fix the package.

Excellent suggestion.

> 
> We might also think about restricting the distribution to a "core" that is
> tested well and then the non essentials of packages available but not
> under the strict release regiment. The distribution is just too big to
> test it all. Perhaps we could have stable/unstable releases only for the
> core release and maintainers of packages outside of the core release have
> their own responsibility of releasing stable/unstable packages whenever
> they wish?

This sounds a little complicated.

> 
> I am very much opposed to restricting releases. If we want the mainstream
> to use Debian then we need regular releases. We need to work out a scheme
> to make this to a routine matter, have establish boundaries of authority
> and most of all simplify the matter as much as possible.

While I agree that the linux world changes quickly enough to
warrent frequent releases, Debian has never put out more than 1
every four months, has it?  That's 3 releases a year -- despite
the desire to do 4.  The latest release didn't go as well as it
could have, so a combination of more realistic goals per release,
a more realistic release schedule, and a better testing mechanism
are in order.  I've heard Bruce address the first of those, and
various people take stabs at the last.  The second (that of
release scheduling) seems to be taking care of itself.

I think Manoj was working on some kind of sorting thingy that
would enable a system to house all of the source packages and
compile the whole shootin' match.  That would be a good test of
the source packages.  The cron job Bruce mentioned will help with
stupid dependency problems for binary packages, so we really only
need to idenit=fy a handful of install profiles and have them
tested.  The distribution is too big for anyone to test the hwole
thing, but it could be parsed out into several general types of
instllations.  The volunteer testors would then only need to test
the various alternatives in each profile.  This sounds a little
confusing even to me at the moment, so I'll try to work on some
specifics.

Thanks.

Richard G. Roberto
richr@bear.com
011-81-3-3437-7967 - Tokyo, Japan


--
*******************************************************************************
Bear Stearns is not responsible for any recommendation, solicitation, offer or
agreement or any information about any transaction, customer account or account
activity contained in this communication.
*******************************************************************************


--
Please respect the confidentiality of material on the debian-private list.
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-private-REQUEST@lists.debian.org . Trouble? e-mail to Bruce@Pixar.com