RE: [GRASS5] Re: GRASS 5.0.1 released

GRASS GRASS developer
1

-----Original Message-----
From: Frank Warmerdam [mailto:warmerdam@pobox.com]
Sent: Wednesday, January 29, 2003 3:21 PM
To: grass5@grass.itc.it
Subject: [GRASS5] Re: GRASS 5.0.1 released

Markus, etc,

Without having any specifics to contribute, I would just like
to second
Markus' point about keeping CVS update access as
straightforward as possible.
As a very occasional committer I have found it difficult to
be sure that I
am working on the right version and have avoided fixing
things myself unless
I am planning to do a fair amount of work.

I'm not even occasional committer. I send this as a start
of a "brainstorming".

Some thoughts:

-There are few contributors.
-Very _few_ people test other people contributions.
-It's difficult to classify a contribution. Grass 5.1 vector
capabilities is a big enhancement, but others (the last I remember:
Paul's modification of *.proj) could be considered anything
between a new feature and a necessary fix depending on who uses it.
-It seems that now there are three versions of grass. One "hyperadvanced"
(e.g. 5.1), one with many enhancements (5.0 HEAD), one "hyperstable" (5.0.1)
-Testing is not proportional to time between releases.
-We have one of the biggest software base with one of the smaller
developer base of the free(open) software world.

And now, my proposal, (remember, just "brainstorming")
-I will propose a shorter release cycle, with not branching. Just
two versions(e.g. Grass 5.1 seem not to be something stable enough).
-CVS would be a "patch monkey", somewhere to contribute code. It
would be great if there were (instead) a human patch monkey.
-One week before releasing a new version, there shouldn't be
contributions, just ¿testing and bug fixes?.
-If a new version goes out with a bug, no problem,
"wait next release".

Hope this helps

Gonzalo

2

On Wed, 29 Jan 2003 GFernandez-Victorio@igae.minhac.es wrote:

And now, my proposal, (remember, just "brainstorming")
-I will propose a shorter release cycle, with not branching. Just
two versions(e.g. Grass 5.1 seem not to be something stable enough).
-CVS would be a "patch monkey", somewhere to contribute code. It
would be great if there were (instead) a human patch monkey.
-One week before releasing a new version, there shouldn't be
contributions, just ¿testing and bug fixes?.
-If a new version goes out with a bug, no problem,
"wait next release".

I like that idea. This seems like a complex problem so a simple solution
probably is the best one. Volunteers with different operating systems
could do a compilation test, and also test the list of changed modules if
possible, in this week with the motivation that their testing would be
incorporated in a release within 7 days.

Having branches seems like overkill for the small amount of people working
on GRASS and the general level of expertise in using CVS. It is a lot to
ask developers to make changes to two branches at the same time and make
sure everything still works. If there were clear advantages to using
branches it might be worth it but it mostly seems to cause confusion and
delays. Although the idea that the code on a release branch can only
possibly become more stable with time is nice, it seems to me that it just
isn't worth it really.

Maybe in the future when GRASS has more developers we could make better
use of branches. But perhaps the time is not yet right for them?

-If a new version goes out with a bug, no problem,
"wait next release".

(or go back to the last one.) That is a nice idea as well.

Just a few thoughts,

Paul