[Standards] reliable messaging

Kevin Smith kevin at kismith.co.uk
Wed Jun 17 11:38:23 UTC 2009


2009/6/17 Remko Tronçon <remko at el-tramo.be>:
> Well, there's not much worse than a reliable messaging protocol that isn't
> 100% reliable. If there's the slightest possibility that this fails (e.g.
> client going offline and coming online with the same resource, and for some
> reason the presence doesn't get through), then it shouldn't be used as a way
> to acknowledge receipt.

When you're at the stage of worrying about broken servers not
distributing stanzas, you've already lost the reliability battle, I
think.

 Message receipts serve a different purpose than transport reliability
though, anyway. Message receipts means that the message has been
'processed' (presented to the user). Using Message Receipts for
reliability, if you couple it with resending when the receipt doesn't
come in within a short time, would result in floods of messages
whenever a user goes AFK (or just is busy with something else).
Message receipts are for user<->user reliability (discussions of
whether that's reliability or not aside), not client<->client
reliability.

/K



More information about the Standards mailing list