On Wed, 21 Jul 2004, [UTF-8] Branko Ä^Libej wrote:
> C. Michael Pilato wrote:
>
> >Branko ÄŒibej <brane@xbc.nu> writes:
> >
> >
> >
> >>I'd like to propose that for 2.0, we _do_ change our internal format
> >>to IRI and let the RA layer handle any conversions. The reason is
> >>simple: there are other clients apart from the command-line client
> >>that don't use svn_opt_args_to_target_array. As it stands for 1.0,
> >>these clients will have to reimplement (or copy our implementation of)
> >>IRI->URI translation, which seems like a bit of a pain.
> >>
> >>
> >
> >Fly. Buick.
> >
> >Just expose the URI<->IRI conversion functions in the API.
> >
> Yuck. We expose too many implementation details as it is. Why should
IRI -> URI is a general purpose function and a documented algorithm, like
much in libsvn_subr.
> ra_local (and ra_svn) convert an URI back to an IRI to to get at the
> path, even though they don't require any escaping? Why should we do this
They will have to anyway, since a URI is an IRI by definition. But there
is no need to do the IRI-AURI for those RA layers anyway.
> conversion in every error message that contains an URI?
>
This is a good question. Depending on the result of this discussion, we
might want to do this everywhere, by a utility function. But if we are
planning to use IRI internally long-term, it is least confusing to give
back what the user entered IMO.
Regards,
//Peter
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Jul 21 21:52:52 2004