On Tue, Mar 18, 2008 at 4:36 AM, Peter Saint-Andre <[EMAIL PROTECTED]> wrote: > > The XSF has been accepted! :)
Great ;) > > > Just for asking if you think it makes sense or not: on the standards > > ml we're discussing several mobile optimizations, and there is always > > the doubt of the real figures behind the solution. What about if we > > opened the source of the xmpp library upon which our client is based > > on and, together with a connection manager (we have also that, already > > opensource) we use a SoC project just for experimenting the solutions? > > My only doubt is that perhaps it's too early for starting coding and > > it's difficult to devise a project with controllable deliverables, the > > only self contained project I've in mind could be a first > > implementation of EXI > > That doesn't sound easy. :) In fact I was just brainstorming. A full EXI implementation on mobile / connection manager would be a kill (in terms of work to do), however if done just at the server side it could be feasible. EXI itself is not so huge, it's just about looking up tokens in dictionaries and doing some reordering, compression is done by standard zlib after all the serialization. The main problem in xml streams is dynamic schema synchronization, but in the the project we could set as minimum goal a parser/serializer with fixed schemas and ASN.1 like encoding of unknown namespaces as EXI already prescribes. I think that such a library (in Java or C), with no client/server integration, could be in the scope of SoC and it would be very useful for both client and server developers in order to start experimenting stream optimizations. -- Fabio Forno, Ph.D. Bluendo srl http://www.bluendo.com jabber id: [EMAIL PROTECTED]
