[Uim] Should we have a stable branch?

[TOKUNAGA Hiroyuki]

On Sun, 09 May 2004 22:55:01 +0900
YamaKen <yamaken at bp.iij4u.or.jp> wrote:

> At Sun, 9 May 2004 03:29:38 +0900,
> tkng at xem.jp wrote:
> > Now I'm thinking whether we should have stable branch or not. Though
> > I'm not so realizing, number of uim users is on the increase, we had
> > better provide a user with the stable software obviously.
> > 
> > Because number of uim developers are not so many, I'm not confident
> > in maintain 2 branches. (Of course we can, but I want to keep
> > development speed as far as possible.)
> 
> I thought about it for a while, but I could not imagine that
> stable branch is effectively working.

I agree with you at this point.

> Inputmethods have many many complex specifications and
> behaviors. Because nobody knows all of this, we have to be
> signaled by scream of actual users when it has been broken.
> 
> This is why we have to release uim frequently. "development
> branch" means "not verified by actual users for a long time" and
> we will have some failures that is difficult to go back.

I don't think so. I'm believing that some users will test development
branch even there's stable release. (So, should we call testing branch
instead of development branch?)

> I suggest an alternative idea. Yes, just an idea.
> 
> 1. release from one branch (as frequent as possible)
> 2. declare released version as "stable" if not screamed for a week

It seems to be an interesting idea. If there's no objection, I will
carry out this idea tentatively. 

Regards,

-- 
TOKUNAGA Hiroyuki
http://kodou.net/