pidgin at unreliablesource.net
Tue Jul 31 11:42:54 EDT 2007
On Sat, Jul 28, 2007 at 01:16:18PM -0700, Sean Egan wrote:
> On 7/28/07, Etan Reisner <pidgin at unreliablesource.net> wrote:
> > No, it isn't hard it just will tend to not be what people expect. The
> > correct way to interpret 'foo at email@example.com' input as a JID is to parse
> > it into Screenname: 'foo' Domain: 'mail at server.com' which is almost
> > decidedly not what the user expected or wanted. But parsing it the other
> > way is broken because @ is not a valid character in the node portion of a
> > JID. Which means that in order for someone to correctly enter that JID
> > they would need to understand how to correctly escape @ in a node (which
> > is to use \40) and put it in as 'foo\40mail at server.com'.
> You're saying that if we tried to parse it,we'd wind up with
> 'foo at mail' as the node and not be able to handle that. How is this at
> all different from the user entering "foo at mail" in the 'node' (or
> 'screenname') box?
No, I'm saying the exact opposite, the person would *want* us to come up
with 'foo at mail' as the node but we would be forced (by definition of the
way JIDs are defined) to parse it as having a node of 'foo' and a server
of 'mail at server.com' (which may or may not be valid in and of itself but
that's a different story). (At least given my understanding of the long
and complicated JID Escaping thread going on at the moment. An
understanding that could very well be entirely wrong as the scope of that
XEP is under somewhat heated debate.)
Which means that to correctly have someone enter a node of 'foo at mail'
would require them to know how to escape the @. Whereas, if we have a
separate node box then we can unambiguously assume that any text input
into the node box is a node and do the correct JID Escaping of the node
part. (Again assuming my understanding is correct about all of this.)
Also note that this is something of a corner case and I brought it up just
so it was out there. I'm mostly still just not sure what confusion we
think we are solving with this change. I am unsure why we think that most
people are going to clearly understand that they need to put in a bare jid
(user at server.com) in one box while we seem to think they have problems
understanding what the relevant parts of the JID even are. Unless I'm
mistaken logging in to gmail doesn't require a foo at gmail.com username, and
I would imagine (though I have never used or seen it) that the Google Talk
client doesn't require that either.
> This doesn't seem like a hard problem for a computer to solve; indeed
> it seems exactly as if it were designed to be solved by computers,
> rather than people.
I never indicated that the problems are hard just that I don't see the
benefit we are trying to attain and that I recall the problems we had when
we tried to guess at adding @hotmail.com to MSN addresses.
More information about the Devel