Re: iSCSI framing using TCP options

[Stephen Bailey <steph@cs.uchicago.edu>]

Ron,

> iSCSI framing using TCP options was raised and noted in an
> unofficial document (Jul/5/2000).  Was this issue ever concluded ?
> If so what is the resolution ?

The TCP option approach was roundly rejected by the IETF transport
community.

The descendent of that effort, which takes the approach of delivering
framing information in-band (in the TCP stream) is still active.  The
authors of that draft are actively working on a new version in
response to many comments on draft-ietf-tsvwg-tcp-ulp-frame-00.txt.

One likely change in the next revision of the draft (which I'm sure
will go straight to last call, and win the hearts and minds of all who
read it :^), is that it will only define what is called PDU alignment
mode in -00.  The other primary client of TCP framing is RDMA, and
that community seems to have no interest in marker mode.  Therefore, I
would propose that iSCSI will continue with its own definition of
marker mode as it currently does, and additionally call out a
negotiation for ULP framing for TCP.  However, I don't think any
specific action is called for until draft-*-01.txt is released in the
next few weeks.

Steph