Re: svn commit: rev 2581 - trunk/subversion/include trunk/subversion/libsvn_wc trunk/subversion/libsvn_subr trunk/subversion/libsvn_client

cmpilato@collab.net wrote:

>=?UTF-8?B?QnJhbmtvIMSMaWJlag==?= <brane@xbc.nu> writes:
>
>
>
>>Greg Stein wrote:
>>
>>
>>
>>>Hmm. Maybe all of our svn_path_uri_* should be apr_uri_*, living in
>>>apr-util.
>>>
>>>Thoughts?
>>>
>>>
>>>
>>>
>>Well of all the ... When I brought this up some time ago, you jumped up
>>and down on me about the uncoolness of using a struct instead of a char*
>>to represent URLs.
>>
>>But +1, I won't change my mind just to get even. :-)
>>
>>
>
>Hm. I've of two minds here.
>
>1. I think it would be a pain to have to tote around URL structures
> where a const char * works just fine.
>
>2. But, at least then we have a bit of type safety that will help to
> keep up in line.
>
>My former mind is currently beating the snot out of my latter one,
>though. *shrug*
>

What I like about apr_uri_parse:

    * It supports all sorts of URL syntax, including
          http://user:passwd@host:port/path
      which we don't support now
    * It makes all the fields of the URL easily available, with defaults

What I hate about apr_uri_parse:

    * It doesn't handle file:// URLs correctly on Windows (something
      Subversion does now, thanks to yours truly), but that can be fixed.
    * The type safety it provides doesn't help where a function can
      accept either a local path or an URL.

If the second hate-point can be resolved, I'm all for moving to struct
pr_uri.

-- 
Brane Čibej   <brane_at_xbc.nu>   http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org