Re: [Gnu-arch-users] Re: today's merges

[Tobias C. Rittweiler]

On Monday, September 1, 2003 at 7:53:24 AM,
    Stephen J. Turnbull <address@hidden> wrote:

[three-branch release process:]
> 1) Stable.  As Tom puts it ``The policy for $STABLE is to fix bugs that
> "have to be fixed" [...]
>
> 2) Development.  Adding features like $FEATURE-X, but requiring that
> they be added in ways that have minimal impact on stable features,
> [...] This limits you to features that don't need low-level changes to
> the core. [...]
>
> 3) Research.  Blue sky, anything goes, for feature design.  [...]

Some not-so-much time ago I figured a very similiar method, but before
having read this, I wasn't really aware of what precisely I had in mind.
As I wouldn't have ever been able to describe it so clearly, I owe you
something.


> This got vetoed, though, on the grounds that nobody would be willing
> to run and work on branch (2), [...]

That makes me wonder. In my opinion about this concept, the development
tree is still the place, where (actually) most of the progress is made.

And so it comes that I wouldn't plaid for _one_ research tree, instead
I'd advocate a loosely coupled network of several trees. Everyone who
wants to try  something different can and should do it---in his own
research branch which he puts publically available and where everyone
can cherry-pick from.
This has the advantage that the range of research keeps openly and wide.

Arch makes it technically possible. :-)



On Monday, September 1, 2003 at 11:07:22 AM, 
    Andrew Suffield <address@hidden> wrote:

> If you have (2), and it's *that* reliable, why bother with (1) at all?

(2) is in no way reliable. After all there gets a damn lot of new stuff
added. It's just that the current infrastructure is adhered to as much
as possible (in contrary to (3)).


-- tcr (address@hidden)  ``Ho chresim'eidos uch ho poll'eidos sophos''