On Fri, Feb 15, 2008 at 8:39 AM, Alexander Gnauck <[EMAIL PROTECTED]> wrote: > > I posted some stats in another thread before: > http://mailman.jabber.org/pipermail/jdev/2006-November/024552.html > > They were not from a very long lived session, but I think they show that > some stanzas compress very well.
The most interesting part is that it's very clear that after a very short transitory initial part, when zlib basically learns the correct symbol dictionary, the compression becomes very effective for the strong correlation between messages. I don't think that the other benchmarks on binary xml reflect well this type of data -- Fabio Forno, Ph.D. Bluendo srl http://www.bluendo.com jabber id: [EMAIL PROTECTED]