Garrett Rooney wrote:
> On 7/15/06, David James <djames@collab.net> wrote:
>> On 7/15/06, Max Bowsher <maxb1@ukf.net> wrote:
>> > In r20677, I deleted a typemap that had been added to the Ruby bindings
>> > since 1.3.x, which caused an undeniable compatibility break, by
>> changing
>> > the return type of svn_repos_get_committed_info().
>>
>> The Ruby bindings don't make any compatibility promises. According to
>> subversion/bindings/swig/INSTALL,
>> "The Ruby bindings API may change incompatibility as development
>> proceeds."
>
> I believe that dates from the time when the initial major work was
> being done on the Ruby bindings, and at the time, it was quite clear
> they they were undergoing major revisions, taking them from
> essentially useless to a quite useful state, but sooner or later you
> do have to say "ok, this is actually released, we need to support
> compatibility". Are we at that point? I personally think we're
> actually past it, but that's just my opinion...
Ditto, and given that we recently revved a Ruby bindings API (time ->
time2) without anyone objecting implies that there there a quite a few
people who are of this opinion.
Is there anyone who would object to defining the Ruby compatibility
policy as:
[[[
Interfaces located within Svn::* are supported and expected to maintain
backward compatibility within Subversion 1.x, EXCEPT for interfaces
located within Svn::Ext::*, which are are internal and subject to change
without notice at any time.
]]]
?
Max.
Received on Mon Jul 17 19:56:28 2006