On 04/03/2013 03:43 PM, Bert Huijben wrote:
> I don't think we really have a hybrid story for keyring, gnome, kde. We
> store everything in there (and don't create a file with pointers). It is
> just Windows where we use the two systems in one.
And I quote (from a file in my ~/.subversion/auth/svn.simple)
{{{
K 8
passtype
V 13
gnome-keyring
K 15
svn:realmstring
V 73
<https://recurrence.googlecode.com:443> Google Code Subversion Repository
K 8
username
V 18
cmpilato_at_gmail.com
END
}}}
>> Personally, I'd like to remove this API before we ship 1.8 and rethink
>> it in light of the work that I plan to revisit post-1.8 with the master
>> passphrase feature. I'm certainly willing to do that (or some other
>> more incrementally improving) work. But I'm not keen on simply
>> jettisoning someone else's code, so I'm seeking feedback here.
>
> I used the current pattern of iterating and allowing to delete from the
> callback to allow incremental improvements in future versions. This moves
> some work to the client that uses this. (Common pattern would be iterate
> to use list and then iterate again to delete items). I would say the
> current api is easier to extend than the 1.0-1.7 'api' of just blowing
> away the directory, which would also leave keyring, gnone and kde hanging
> around.
I'm not arguing that your programmatic approach isn't better than "blowing
away the directory". I just don't want to get stuck with an API that pins
us down. The more I think about the interfaces you've exposed -- ignoring
the details of the implementation -- the more I believe they will be okay.
There's still plenty of stuff that, IMO, must be fixed before 1.8.0, but
I'll withdraw the suggesting to drop the API. I know TortoiseSVN (and
others) want it.
--
C. Michael Pilato <cmpilato_at_collab.net>
CollabNet <> www.collab.net <> Enterprise Cloud Development
Received on 2013-04-03 22:12:05 CEST