- From: Laurent Eschenauer <laurent@eschenauer.be>
- Date: Mon, 17 Sep 2012 21:40:51 +0200
- To: Evan Prodromou <evan@status.net>
- Cc: public-fedsocweb@w3.org
- Message-ID: <CAO9NFWrY1m-i+6GTX6ELrJ32ijsv3jgUX1wRedfBGP_3MK_sZw@mail.gmail.com>
Hi Evan, My 2 cents on XMPP vs HTTP vs WS > > - XMPP comes with a mature underlying model for endpoints (accounts, > servers, components) and payloads (messages, ims, presences). That's great > if it maps onto what you want to do... otherwise you have to adjust. > Websockets is much more like a bi-directional TCP/IP connection. It > has practically no model whatsoever. > > I don't understand this one. XMPP brings you federation, with a proven identity and security model, many highly scalable implementations, etc. You'll have to re-invent all of this in WS world. I'm still convinced that XMPP would the fastest and most reliable way to bring this federated social future we dream of. The only reason to go HTTP (and re-invent things like PubSub, Dialback etc over HTTP) is developer reach and infrastructure maturity. There are many more people out there who understand HTTP than XMPP and we've learn for 30 years how to build highly scalable HTTP services. > > - For non-XML payloads, e.g. JSON, you end up doing a lot of > XML-wrapping with XMPP. With Websockets, you're free to send just about > anything down the pipe. > > All depends on how you go about it. You can leave the XML work to the server and undelying lib and just focus on the payloads. I don't see why these could not be JSON messages. Replacing salmon by xmpp messages for cross-domain notifications would be a good start. > > > I think there's room for experimenting with both mechanisms. > Of course, keeping an open mind, I like the fact that WS could be used both at S2S and C2S level. Unfortunately XMPP is missing in the browser and has to resort to tricks like BOSH. Experimenting is fun and full of learning. However, in the end, you'll have to reinvent everything XMPP already has (identity, discovery, authentication, etc etc). Best, Laurent
Received on Monday, 17 September 2012 19:41:39 UTC