|
I agree. The *most* that we might have to change is to make
it clear that an
iWARP implementation *may* choose to interoperate with the
RDMAC versions,
and that there is nothing in the main drafts that forbids
this.
The requirements on what is a "valid" version number is the
only thing that
might have to be clarified, so I think it is doable. I also
do not think it is
necessary, since interoperability is extra functionality
which has no impact
on the pure IETF protocol. But it is better to be clear
that more liberal interpretation
of "valid" that enables interoperability *is* legitimate
and even ever so slightly
encouraged (but probably not to the level of a
"should").
I am very concerned that the
current path we’re on is to try to do last call before we close on the interop
issues with the RDMA Consortium protocols. To me this would be a mistake
because it effectively fragments the market when in fact we have a very
concrete interop proposal on the table that appears to be holding water, which
has very specific text change recommendations that appear to be
straightforward to put into the specs.
Can we quickly get through the
interop changes to MPA/DDP/RDMAP? Seems like the only outstanding issue raised
so far is what is the exact text used in the RDMAP/DDP specs on what is a
“valid” version number (i.e. do we leave it up to the vendor, or do we state
something more explicit). That seems like a pretty small outstanding issue?
Further, it seems like we should
be able to get out updated drafts within a few days which incorporate the
recommendations, and the recommendations do not have significant complexity to
implement in the normative body of the specs (unlike the outstanding issue of
normative statements in the security chapters of DDP/RDMAP
(!!).
Jim
|