Lars,

> have the authors verified that -05 addresses the concerns of 
> the TCPM reviewers?

The answer is "not completely" because the WG chair (yours truly)
has been busy behind the scenes, and has a plan, but forgot to
explain the plan to the WG and AD ... bad WG chair ... mea
culpa.

My plan is that this is the first of two revisions of the draft;
the primary purpose of this revision is to provide a basis for
discussion at TCPM in Montreal.  After that discussion, there will
be another revision to take care of the results.

The key change that is happening in the MPA draft is summarized in
this sentence from the summary of changes in the -05 version:

        Document restructuring to differentiate between fully
        layered MPA on TCP implementations and optimized MPA/TCP
        implementations. 

In looking at the primary TCPM review, it became apparent that a
significant source of the issues is that MPA is intended to be
implemented with TCP with massive cross-layer optimizations so that
the combined optimized MPA/TCP (plus other RDDP protocols) achieves
aligned zero copy on receive and the protocol is specified as if
those optimizations are always present.  Think of the optimizations
as putting MPA and TCP into a single functional box, heavily optimizing
the internals of that box, while preserving all external interfaces,
including the socket interface for other applications - see Figures
7 and 8 in the draft for one version of the before and after.

The result of these optimizations is that what was the socket
functional interface between MPA and TCP gets seriously changed
to make aligned zero-copy receive work.  The older MPA drafts
documented these changes *as if they were changes to the TCP
interface* and hence would be usable by other applications/protocols.
>From the viewpoint of the sockets interface and TCP functionality,
this approach creates a number of problems underlying the review
comments.

The direction that the -05 draft is headed in is to specify MPA
as a cleanly layered protocol on top of TCP (can use the existing
sockets interface) whose straightforward implementation results
in a copy in the RDDP/TCP receive path.  Then, the combination
of MPA and TCP is put in a box, and the internals of that box are
optimized for zero copy without change to the external interfaces
(if there are other apps, they still use sockets as they exist);
the -05 draft describes the key things that need to be done to get
the result to be zero-copy, and is intended to do so in a fashion
that has no impact on the sockets interface for other protocols.
This should result in an absence of statements that appear to make
normative changes to TCP's socket interface to applications.

The intent here is to discuss this further at TCPM in Montreal.

Thanks,
--David
----------------------------------------------------
David L. Black, Senior Technologist
EMC Corporation, 176 South St., Hopkinton, MA  01748
+1 (508) 293-7953             FAX: +1 (508) 293-7786
black_david@emc.com        Mobile: +1 (978) 394-7754
----------------------------------------------------

> -----Original Message-----
> From: Lars Eggert [mailto:lars.eggert@netlab.nec.de] 
> Sent: Wednesday, June 28, 2006 12:28 PM
> To: rddp-chairs@tools.ietf.org
> Cc: rddp@ietf.org
> Subject: Re: [rddp] I-D ACTION:draft-ietf-rddp-mpa-05.txt 
> 
> Hi,
> 
> On Jun 27, 2006, at 21:50, Internet-Drafts@ietf.org wrote:
> > A New Internet-Draft is available from the on-line Internet-Drafts  
> > directories.
> > This draft is a work item of the Remote Direct Data Placement  
> > Working Group of the IETF.
> >
> > 	Title		: Marker PDU Aligned Framing for TCP 
> Specification
> > 	Author(s)	: P. Culley, et al.
> > 	Filename	: draft-ietf-rddp-mpa-05.txt,.pdf
> > 	Pages		: 72
> > 	Date		: 2006-6-27
> 
> have the authors verified that -05 addresses the concerns of 
> the TCPM  
> reviewers?
> 
> Thanks,
> Lars
> -- 
> Lars Eggert                                     NEC Network 
> Laboratories
> 
> 
> 

_______________________________________________
rddp mailing list
rddp@ietf.org
https://www1.ietf.org/mailman/listinfo/rddp