[jdev] Re: Public Server with Compression Available

[Hal Rottenberg]

What clients and libs are working on this?

On 1/10/06, Stephen Pendleton <spendleton at movsoftware.com> wrote:
> I can also confirm that it works well: I am seeing reductions of the XML
> stream of about 60%. Does Wildfire have compression for s2s connections as
> well as c2s?
>
> Stephen
>
> -----Original Message-----
> From: jdev-bounces at jabber.org [mailto:jdev-bounces at jabber.org] On Behalf Of
> JD Conley
> Sent: Monday, January 09, 2006 9:12 PM
> To: Jabber software development list
> Subject: RE: [jdev] Public Server with Compression Available
>
>
> I thought I'd fire off a little update. I've heard from four different
> client project developers. They were all able to get zlib based compression
> working with only a few little stumbling blocks. There are some truly
> impressive compression ratios being generated over time! I'll let them
> divulge that information, though.
>
> Here were some of our stumbling blocks:
>
> 1) If you have control over the zlib header and checksum footer for DEFLATE
> (only supported by some libraries) you need to make sure you include them.
>
> 2) Make sure you flush your deflater instance so it completely compresses
> the input buffer. If you're using the reference implementation
> (http://www.zlib.net/) this is done by passing a Z_SYNC_FLUSH. You don't
> want to do a Z_FINISH, as you should preserve the dictionary for the entire
> XMPP stream for better compression.
>
> 3) Keep an instance of your inflater and deflater around for each socket
> connection.
>
> 4) Don't assume that one read from a socket will yield you something that
> can be inflated. Also don't assume once you get an inflated buffer that it
> will be a stanza.
>
> If anybody else wants to test on beta.soapbox.net, feel free. The server is
> still running (and has S2S enabled now).
>
> -JD Conley
>
>
>


--
Psi webmaster (http://psi-im.org)
im:hal at jabber.rocks.cc
http://halr9000.com