> On of the current things with WiaB seems to be that the server and client are > all rolled in to one. For people wanting desktop clients, mobile clients, > etc, I think it is critical to have a well defined protocol for the client > and server to work together. > > ~Michael
Couldn't possibly agree more with that! It would be nice if client development could be done relatively independently from server development, I think both would proceed faster if they wernt so tied together. -Thomas > > On Apr 9, 2011, at 3:32 PM, Thomas Wrobel wrote: > >> Remember there is strong pro XMPP voices here too; >> http://www.process-one.net/en/blogs/article/xwave_a_tribute_to_google_wave_team/ >> >> I'm not sure who is right on a technical sense, but there is working >> xmpp based federations out there, are their any based on http? >> >> Regarding the c/s protocol - it makes somewhat sense anyway if it was >> more or less the same as the server/server one, seeing as they both >> basically have to exchange the same information to keep the document >> in sync. >> >> As I suggested on the recent poll,however, I think separating out the >> wiab webclient and using a lib for c/s protocol would help a lot. That >> way , even if different choices are made for the protocol later on, >> its just the lib that has to be changed. People could thus make >> clients with the protocol itself abstracted away. >> >> ~~~~~~ >> Reviews of anything, by anyone; >> www.rateoholic.co.uk >> Please try out my new site and give feedback :) >> >> >> >> On 10 April 2011 00:13, Michael MacFadden <[email protected]> >> wrote: >>> Nelson, >>> >>> There has been large debate on XMPP in wave. The general complaint is that >>> the protocol is to verbose. My two cents are that one of the main points >>> of XMPP was a verbose XML human readable protocol with standard extension >>> mechanisms. However wave uses protobufs and base64 encodes all the data in >>> the XMPP stanzas. The data exchanged by wave is not human readable, xml >>> based, or part of the XMPP standard. That defeats the purpose of using the >>> XMPP standard in the first place. In my opinion this basically relegates >>> XMPP to just a delivery envelope, and one that adds on a lot of overhead. >>> >>> Also XMPP's dependance on long lived TCP connections to maintain the xml >>> stream, there are difficulties providing services to clients that are >>> frequently disconnected. For these reasons there is talk of adding a "raw" >>> http transfer mechanism for federation. Until that is worked out I would >>> hesitate to entertain the idea of injecting XMPP in to the c/s protocol. >>> >>> ~Michael >>> >>> On Apr 9, 2011, at 2:51 PM, Nelson Silva wrote: >>> >>>> XMPP over websockets was proposed as a draft >>>> (http://tools.ietf.org/html/draft-moffitt-xmpp-over-websocket-00) and >>>> ejabberd now has a sample module to support it: >>>> >>>> http://blog.superfeedr.com/xmpp-over-websockets/ >>>> >>>> Wouldn't it be great to use XMPP for both C/S and federation ? or is it >>>> too verbose ? >>>> >>>> Just wanted to share this with the list. >>>> >>>> Regards, >>>> >>>> Nelson >>> >>> > >
