On Fri, Mar 28, 2008 at 6:46 PM, Peter Saint-Andre <[EMAIL PROTECTED]> wrote:
Besides Justin's observation, which is correct, which are the possible problems of BOSH for moobiles? I can just think of one, bandwidth due to the additional HTTP headers and the <body/> bag. If so it's more reasonable defining some "binary" (no joke intended... ;)) headers containing contenght length, sid and rid, than adding incredible complexity by switching transports. We would lose the ability of passing through web servers, but I don't think that specialized BBOSH (Binary BBOSH) would be a problem, if the real issue is bandwidth. Another possible optimization could be allowing compression on the raw sockets in which we pipeline HTTP requests, thus compressing headers too (that is what normally happens with HTTPS...). > > So a few questions: > > 1. Is BOSH acceptable in practice for high-activity periods? > > 2. Do people think that this kind of transport-switch will solve > existing problems? > > 3. Would the costs (complexity of switching) outweigh the benefits? > > I think we might want to consider something like this in the future if > we discover that BOSH doesn't work well in practice, but IMHO we need > more deployment experience with BOSH before we can know the answer. -- Fabio Forno, Ph.D. Bluendo srl http://www.bluendo.com jabber id: [EMAIL PROTECTED]
