Re: [gentoo-dev] Round 2: GLEP 19 -- Gentoo Stable Portage Tree

[Daniel Robbins]

On Tue, 2004-02-03 at 14:39, Paul de Vrieze wrote:
> Basically when one maintains a farm of computers with many users that use it
> for various purposes there are a number of issues at play:
> - New versions could introduce new bugs that some of the users might hit
>   (going back is often a problem)
> - New versions could remove or change features that particular users want. In
>   any case any non-bug-fix release will create some level of user confusion
> - Install's are often image based. While it is possible to have a few changes
>   be propagated after mirror installation, bigger changes need to be included
>   on new images with all the needed testing etc.
> - Many company policies demand new rounds of testing before a new version of
>   any package is released. The smaller the change (security fix only, usually
>   a patch of less than 30 lines), the less testing is needed.
> - Each and every change often needs to be manually reviewed. If there is just
>   a security patch there will be no changed dependencies and less effort
>   needed for the review.

I think this is a good start on documenting the various issues that we
are trying to address. However, I think we need *specific* examples of
these issues so we can analyze them and make sure that whatever solution
that is chosen is indeed the best one. This all needs to be documented
in the GLEP so that we all understand exactly what problems we are
trying to solve. The goal must be clear.

The steps we should follow are:

1) document the specific issues (real examples experienced by real
people) that we are trying to address. Document expectations that users
have for Gentoo in a production environment.

2) define requirements that need to be met for any implementation.

3) have different implementations proposed by developers, users. Have
these proposals reviewed and commented on, possibly refined.

4) choose the implementation that solves the problems and best meets the
requirements.

5) develop an implementation roadmap

6) implement the solution

People are jumping to step #4, but I don't think that we've even
completed step #1 yet... certainly not to the level that it needs to be
in order for us to make sure we're making the best decision. 

And remember, after step #1 comes #2, defining requirements.... *then*
step 3: proposed implementations.

Long live nerdboy!

Regards,

Daniel


--
gentoo-dev@g.o mailing list