Re: Continuing discussion of oaf ...

[Martin Baulig <martin home-of-linux org>]

Miguel de Icaza <miguel helixcode com> writes:

> For instance, say I have Gnumeric that implements the
> `IDL:GNOME/Gnumeric/Workbook:1.0' interface, the OAFID of this guy
> could be `GNOME/Gnumeric/Workbook:1.0' (note that there is no IDL in
> the prefix).

I still think it's a bad idea to name it this way - we shouldn't do things
in a way that they work for us but will be wrong if you do the same in your
application.

People learn from copying stuff and your proposal gives people the wrong
impression that they can just use the IDL name as the OAFID.

> Now.  I believe the issue of uniqueness is not going to come up if we
> have a few guidelines for constructing the OAFID name space (and still
> manage to not use the ugly uuids).

I'm wondering since when uglyness has become an argument against technical
requirements.

> > It's a lot easier for two developers working completely independently
> > to accidentally pick the same human-readable string name than to
> > accidentally pick the same UUID; the latter is more or less impossible
> > short of doing it intentionally.
> 
> But this will hardly ever happen.  There are few binaries that are
> called /usr/bin/ls.  By being more verbose on the name we can avoid
> this issue completely (for instance, by saying that the org/gnome
> registry would be managed by the GNOME foundation).

Using UUIDs will just totally avoid that we will ever get this problem - and
as I already proposed in that msg00179.html, I proposed that it should be
allowed to use the short name in queries while the UUID is still required.

Btw. /usr/bin/ls is really a bad example - you should have choosen something
like a text editor, an image viewer, or an IRC client or something like this
which people will actually want to write.

-- 
Martin Baulig
martin gnome org (private)
baulig suse de (work)