Or maybe 3 repositories: kernel, libs, all

[Brian Rice]

Sorry for not replying sooner. I had to think about it a bit.

On Apr 1, 2005, at 5:24 PM, Paul Dufresne wrote:

> I was just thinking about my previous messages about eliminating alpha.
>
> I was thinking that somehow, main is a kind of experimental repository.
> My previous idea had the problem, that you would have to plan much more
> the release. That is: think of new design stuff for next stable 
> release,
> add it to the repository, change the consequences of this in the
> libraries, and then in the programs that use them, then make a new 
> Alpha release and rebegin this cycle. And if we would be about to 
> release Alpha, new
> touchy patches would have to wait.

What? The whole idea of main is that we can put all sorts of patches in.

> But part of my problem for current situation, is that, I would prefer
> not to be expose to main changes, and continue my work like if there 
> was no
> important changes going on. But almost everybody use main, so I can
> hardly send SDL patches for alpha only, without them to go in main. I 
> have the
> feeling we have the philosophy that main should contain everything, and
> alpha, almost everything.

I realize it's a problem for you to work with main, but if you don't, 
then I won't be able to perform patch merges. No one else edits your 
SDL code, so why don't you just edit in another directory that's 
insulated from main, and then copy the files over and then work with 
darcs?

That's my suggestion to make it easier for you: maintain a couple of 
different branches, and just copy over SDL changes before performing 
darcs record; or maybe afterwords. You DON'T have to work in the same 
directory as you post, for the very reason you describe: your code is 
not sensitive to other code, nor vice versa.

> Now, what just came to my mind, would be to have 3 repositories:
>
> kernel would contains just files for generating a new Image and a new
> VM.
> libs would contains src/lib repositories, and other almost essential
> stuff.
> all would contains everything.
>
> kernel is very experimental.
> libs is lightly experimental.
> all is somehow stable.
>
> Now, as a developper of SDL, I would probably just work on all, and 
> send
> my patches against the all repository.
>
> Once in a while, there would be an email saying: we are about to 
> release
> the libraries, please test your code against it before we apply libs 
> on all.
> I would then test SDL code against latest libs repository, and give 
> feedback on how well SDL works against the latest libs.

Considering that the maintenance of this stuff comes down to one 
person, ME, and that the current setup is barely within my control, I 
will have to refuse 3 layers until we have more people and volunteers 
or a more stable code-base, and the stability is not likely for a long 
time.

> A similar process would go on between kernel people, and libs
> developpers, but I would probably not care about it.
>
> I really like to have feedback on this idea.

Sorry, but this is just going to have to wait or not happen at all. 
Just change your development practices slightly, and it will be 
smoother.

--
Brian T. Rice
LOGOS Research and Development
http://tunes.org/~water/