The IP over ATM Mailing List Archive by date

["Eric W. Gray" <egret@bluefin.net>]

Raif Onvural wrote:
> 
> *** Resending note of 03/28/96 18:12
>  - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
>  From:
>  Subject:
> Subject: Mpt-Pt AAL5 delimiting
> Bert, how does this differ from making those points VP end-points,
> have end-to-end connections cross one or more VP links,
> as needed and terminate at the VP end points as needed and use
> VCI for demultiplexing. This would make it all legal :). In either
> case, there is the problem of allocating unique VCs/source.
> Raif
> xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> If AAL5 cells can't be interleaved because there's no way of unscrambling
> them after the fact,
> 
> If there would tend to be few senders and many more receivers in a mpt-pt
> connection, and
> 
> If one wants to establish one VC for the mpt-pt connection, then
> 
> I wonder about a bastardized use of VPIs? How about each sender uses the
> same VCI but different VPIs? This would allow interleaving of AAL5 cells
> at all merge points while still keeping them identifiable. And scaling
> might not be a real problem.
> 
> This strange use of VPIs could become a problem when the intended use of
> VPs is implemented, but then again ATM switches could manage both uses of
> VPs separately, no? Couldn't switches dynamically keep track of which
> VPIs were to be terminated at own switch for mpt-pt connections, and
> which VPIs were just for cells in transit?
> 
> Somehow, the meticulous buffering of n packets' worth of cells along the
> way is sure to incite the ire of the ATM god.
> 
> Bert
> manfredi@engr05.comsys.rockwell.com
> 

Either way, you don't end up with a single VC - the VPI/VCI combination is
how switches tell one VC from another.  Even if you could get all switch
makers to go along with such a scheme, it doesn't save that much on over-
heads associated with using separate pt-pt VCs.

--
Eric Gray