RE: first class support for flow/correlation/reply-with header co ncept?

Sanjiva,

Session correlation of longer running conversations is absolutely needed in
many scenarios, but it usually goes along with other requirements such as
ordered delivery of messages.  It seems to me that this would require a
separate protocol layer on top of XMLP.

Satish

-----Original Message-----
From: Williams, Stuart [mailto:skw@hplb.hpl.hp.com] 
Sent: Tuesday, February 20, 2001 7:21 AM
To: 'Sanjiva Weerawarana'
Cc: xml-dist-app@w3.org
Subject: RE: first class support for flow/correlation/reply-with header co
ncept?


Sanjiva,

The current abstract model draft [1] models the service provided by XMLP as
supporing both a one-way operation and a two-way request response operation.
The implication of the latter is very much the correlation of the 'request'
and 'response' messages from a single operation that you seem to asking for.
The mechanism that provides such correlation within the protocol remains a
matter of design. The model does not model more conversational/session
oriented exchanges. We would have to do more work to abstract longer running
exchanges or conversations that have a free-er message exchange pattern. 

[1] is a work in progress, so does not represent a concensus point amongst
the members of the WG.

Regards

Stuart Williams
[1]
http://www.w3.org/2000/xp/Group/1/02/16-abstract-model/XMLProtocolAMG.html

> -----Original Message-----
> From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
> Sent: 20 February 2001 12:13
> To: xml-dist-app@w3.org
> Subject: first class support for flow/correlation/reply-with header 
> concept?
> 
> 
> IMO it would be useful if XMLP were to support some first
> class notion of
> flow ID (or correlation ID) as a header. Another way would be 
> to have the
> ability to tag a header as some kind of "reply-with". 
> 
> This would allow one to use XMLP to invoke session-oriented
> services even
> when carried over non-session-oriented protocols.
> 
> Is there any intent to support this capability (or an equivalent)?
> 
> Thanks,
> 
> Sanjiva.
>

Received on Tuesday, 20 February 2001 12:54:28 UTC