Re: [Simple] Message Stateful MSRP Switches (was Re: Multi-party chatWGLC?)

[Ben Campbell <ben at estacado.net>]

On Oct 13, 2009, at 5:03 AM, Geir Arne Sandbakken wrote:

> Ben,
>
> It will be hard to avoid being stateful. The headers of the
> "message/cpim" wrapper might be split in two chunks.  The switch  
> must be
> stateful until a full header set is received. The chunk example in -05
> shows a split in the headers of the "message/cpim" wrapper.

Just to be clear--I agree that it will be hard not to be stateful. I  
didn't mean to object to the change--I'm just trying to make sure  
everyone notices and understands it.


> Thanks,
> Geir Arne
>
> -----Original Message-----
> From: simple-bounces at ietf.org [mailto:simple-bounces at ietf.org] On  
> Behalf
> Of Ben Campbell
> Sent: 13. oktober 2009 00:05
> To: Geir Arne Sandbakken
> Cc: simple at ietf.org
> Subject: [Simple] Message Stateful MSRP Switches (was Re: Multi-party
> chatWGLC?)
>
> <as individual>
>
> I'm still in the process of reviewing this version, but I want to
> point out one particular change that went in as a result of the
> previous WGLC feedback:
>
> Both versions rely on "message/cpim" metatdata for certain MSRP
> message routing and authorization decisions. The previous version
> overlooked the possibility of a large message being sent as a series
> of chunks. Chunking occurs after MIME encoding. When a message/cpim
> document gets chunked, the metadata will only be present in the first
> chunk. Revision 4 assumed that each MSRP send request would include
> the metatdata.
>
> Revision 5 deals with this by requiring the MSRP switch to be message
> stateful. It has to remember the message/cpim metadata until it
> processes all chunks of a given message. This metadata "state" is
> indexed by the MSRP Message-ID header field., which is the same for
> all chunks of a given message.
>
> IMO, this is the best approach of any that I have heard of so far. It
> is unfortunate that we need to keep message state, but I don't see an
> alternative short of forbidding chunking of messages to an MSRP  
> switch.
>
> What do other people think about this?
>
> Thanks!
>
> Ben.
>
>
>
> On Oct 9, 2009, at 3:39 AM, Geir Arne Sandbakken wrote:
>
> > Experts,
> >
> > The 05 version resolves the issues raised in the first WGLC of the
> > draft.  Does the WG think it is ready for another WGLC?
> >
> > Thanks,
> > Geir Arne
> > _______________________________________________
> > Simple mailing list
> > Simple at ietf.org
> > https://www.ietf.org/mailman/listinfo/simple
>
> _______________________________________________
> Simple mailing list
> Simple at ietf.org
> https://www.ietf.org/mailman/listinfo/simple
> _______________________________________________
> Simple mailing list
> Simple at ietf.org
> https://www.ietf.org/mailman/listinfo/simple