2007/6/18, Xavier Hanin <[EMAIL PROTECTED]>:
On 6/18/07, Gilles Scokart <[EMAIL PROTECTED]> wrote:
>
> I agree with this :
> 2.0-alpha-2 : (bug fix + code cleaning) early-july
> 2.0-beta-1: (bug fix + code cleaning + tutorial) late august
> 2.0-RC1 (all major bug fixed + code cleaned + tutorial/doc updated)
> late septembre
> 2.0-RCx (all major bug fixed) every 2 weeks
> 2.0 final : october/november (why not exactly 1 year after the vote
> that accepted ivy in the incubator project)
>
> However, I'm still thinking that we should give us more time before
> publishing an API. I think we could draft it, collect feedback from
> it, but not yet release it. My prefference would be to do a complete
> API in a 3.0.
I think it's a matter of words. I would like to get in the 2.0 version at
least a statement saying what is considered public in the API, and thus for
which we will maintain backward compatiblity in 2.x versions. Even if this
is a very minimal scope. To get a clean and reviewed API, I think we agree
that we'd need to call it a 3.0 version. So I see no problem with breaking
even things we'd agree to say it's public in 2.0. But I think being able to
at least run a resolve from the API with confidence of stability for
the 2.xstream is not a huge work and would help some users.
I see your point. Having a 2.0 also usable as a library with a
minimal API would be nice. However, I'm not sure that offering
something and saying "it's minimal, and we already have in mind to
change it in 3.0" is a good idea. (I exagerate maybe a litle bit, but
that's the impression that I have).
What does the other people on this list think.
--
Gilles SCOKART
