On Mon, Mar 31, 2008 at 7:29 PM, Justin Karneges
<[EMAIL PROTECTED]> wrote:
>  The current situation is a mess.  While XEP-198 has a high XEP number, the
>  concept is many years old, and when it was first introduced there was little
>  interest and the council rejected the proposal.  It didn't see the light
>  until seventy-four XEPs after BOSH, and during that period developers
>  realized that maintaining TCP connectivity can be a problem and that BOSH
>  solves the problem.  Complete disaster.  Now there is interest in promoting
>  HTTP as the best transport for XMPP?  How in the holy hell did this
>  happen? :)

It's not about HTTP, it's about a protocol which puts boundaries to
the messages. If you ever programmed in environments where the
connection is unreliable, raw TCP is a nightmare because it does
exactly the opposite of what you'd like: detect as soon as possible
that the connection is broken and try to understand how many bytes
you've been able to send.
What I'd prefer for mobile clients is a TCP transport with this train
of fields (xmlstanza may be empty if it's just an ack):
*seqid*length*xmlstanza*

(the first message upon reconnects should contain also a session id)

bye

-- 
Fabio Forno, Ph.D.
Bluendo srl http://www.bluendo.com
jabber id: [EMAIL PROTECTED]

Reply via email to