Hello. I'm looking over the code base and like I said before, I think it looks great.

I have a question that I can't answer for myself just yet (since I'm not a mina expert it's hard for me to digest all of the Io/Codec/Protocol handlers). The question is, how hard could one connection support multiple resources or SessionContexts associated with it.

In XEP 0193 (incorporated into RFC3920bis), it says that from one steam-connection, the client is allowed to bind to multiple resources/sessions. And then in XEP 0225, they use this to have external components bind to the server (instead of the old XEP 0114), and I believe this could be used by federated servers too (not sure if that is mentioned in XEP or not). (urls are below)

Since Vysper doesn't have XEP0114 yet nor really support federated servers yet, I would vote to go straight into supporting 0193/0225. I think this would make vysper more powerful in the long run, as well as having a cleaner more consistent code base...

The second reason I like the ideas of 0193/0225 is that this could really help for implementing clustering components. I can see how a component could connect with two JIDs, one the component domain (component.domain.com), and a second would be a component-node-jid ([email protected]/#####). So that components could communicate with each other using their component-node-jid as a back-channel for coordination or what not, while still handling normal component traffic to their full domain.

So I'm sorry, at the moment I'm not the one who can implement this ( i'm not a mina expert yet ). But I wanted to get the idea out there before someone starts to implement external components. :)


http://xmpp.org/extensions/xep-0193.html
(incorporated into RFC 3920bis http://tools.ietf.org/html/draft-saintandre-rfc3920bis

http://xmpp.org/extensions/xep-0225.html


<digression>Anyone located in SF/BayArea? Maybe people want to meet face to face?</digression>

Reply via email to