On 2/4/10 2:34 AM, Kirk Bateman wrote:
> Inline comment below
> 
> On 4 Feb 2010, at 04:22, Peter Saint-Andre <[email protected]> wrote:
> 
>> On 2/3/10 9:07 PM, Liam wrote:
>>>> Understood. I'm going to chat about this with Jack Moffitt at the XMPP
>>>> Summit next week to see if he has any insights. I'm not a web guru. :)
>>>
>>> Tell him I said hi (I've been working on the strophe pubsub plugin a lot
>>> recently :)
>>
>> Will do. :)
>>
>>> I improved my browser-tab-close issue by ensuring that strophe issues a
>>> disconnect on tab-close. However there is no way to know that a stanza
>>> which tcp transfers ok has been processed by the client without an
>>> explicit ack mechanism. Closing a tab, etc. won't close the socket right
>>> away, or even stop reading it.
>>
>> Right. Did you read XEP-0184?
>>
>>> The nature of "publish-subscribe" implies greater reliability than IM,
>>> to my mind.
>>
>> Not to mine. I care more about a personal IM conversation than some
>> random syndicated data. :) But YMMV.
>>
> Peter, how about if xmpp was being used for financial transactions. I'd
> probably use pubsub for that as a queue type thing ( as mq systems are
> often used currently ). Surely that needs a bit more in the way of
> guaranteed delivery ? Just a thought.

Those weren't the original use cases for XMPP. We've slowly built more
reliability into XMPP, but *guaranteed* delivery (like "security") is
not something you just sprinkle on top. :) Even, say, XEP-0184 for
end-to-end acks doesn't *guarantee* that a message was properly handled
at the receiving end or shown to a human user (only that the receiving
application got the stanza).

Peter

-- 
Peter Saint-Andre
https://stpeter.im/



Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to