Ralph Meijer wrote:
> On Wed, Dec 08, 2004 at 02:26:30PM -0700, Peter Saint-Andre wrote:
>> In article <[EMAIL PROTECTED]>,
>>  Ralph Meijer <[EMAIL PROTECTED]> wrote:
>>
>>> On Wed, Nov 17, 2004 at 10:02:40AM -0700, Peter Saint-Andre wrote:
>>>> Two further issues/questions:
>>>>
>>>> (1) The more I look at, the more I think that it would be best to 
>>>> encapsulate disco/pubsub NodeIDs in the query component, e.g., something 
>>>> like ?disco&node=foo. This seems simpler than putting it in the path, 
>>>> which strikes me as unnecessarily complicated.
>>> Yes, I agree. In XMPP, the node is also 'just' an attribute of some specific
>>> action (like disco).
>>>
>>> For the moment we can leave out path segment parameters entirely, but maybe
>>> we should reserve the allowed reserved characters for future use? Did that
>>> parse?
>> Hmm, but designing for unknown use cases is probably not a good idea.
> 
> Well, after a bit over three years [*], I am finding me disagreeing with
> myself.

Wow, I think you have broken my record for longest time between post and
reply. :)

> Within the work on federating social networks, 

Well, now you are designing for known use cases instead of unknown use
cases. It makes a difference. :)

> I am looking at linking
> to publish-subscribe nodes from HTML and Atom documents like so:
> 
>   <link rel='xmpp' href='xmpp:...'/>
> 
> The rel 'xmpp' is debatable, but as far as I know there is no one
> referring to XMPP entities from HTML yet. In this case it could refer
> to any kind of xmpp entity: users (' accounts), MUC rooms, servers, etc.

Right. The rel='xmpp' seems fine to me. Is there a registry of rel
values? I don't think so.

> The problem I have is that there is no nice way to encapsulate the
> combination JID + NodeID in a URI without specifying an action.

Correct. That's because we never went all the way with nodes by making
them part of the address or considering a node to be an addressable entity.

> I would like to avoid specifying an action because I believe it is
> better if a consumer of such an URI would maybe use service discovery on
> the JID/NodeID and then provide possible actions based on that. For
> pubsub nodes, that might be subscribing, but also maybe retrieving
> previously published items.

So what about this?

   xmpp:[EMAIL PROTECTED];type=get;request=info;\
                     node=f9451384-e23e-449f-8a8b-918649205f3c

> So I am tempted to come back to my original proposal of using URI Path
> Component parameters looking like this:
> 
>     xmpp:[EMAIL PROTECTED];node=f9451384-e23e-449f-8a8b-918649205f3c
> 
> Opinions?

Basically that makes the NodeID part of the address, because rfc4622bis
says "The path component of an XMPP IRI/URI identifies an XMPP address
or specifies the XMPP address to which an XML stanza shall be directed
at the end of IRI/URI processing."

I need to look at rfc4622bis, RFC 3986, and RFC 3987 a bit more to know
what I think about this, but my immediate reaction is that it's not evil.

> I'm cross posting to [email protected] because I'm not sure if anyone reads
> this list anymore.

I'm dropping [EMAIL PROTECTED] because we don't use that list anymore.

Peter

-- 
Peter Saint-Andre
https://stpeter.im/

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to