Re: iSCSI: Towards Urgent Pointer Consensus

[Glen Turner <glen.turner@aarnet.edu.au>]

"Randall R. Stewart" wrote:
> 
> I don't think this group has any license to change TCP API or other
> parts of TCP for that matter...
> 
> I think it is much more fundamental than the API. Has I have stated,
> after looking at two implementations, you are defining a
> behavior that IS NOT assured. When you use the URG pointer as
> a record marker under multiple loss conditions it WILL FAIL to do
> what is being defined.

Hi Randall,

These are points that I have been making for some time and were
apparent about 10 minutes after I started coding a user-space
target for generic UNIX platforms (so I'd have something to test
a kernel-space Linux implementation against -- I won't be able
to afford a 'real' iSCSI target disk).

I've also been at pains to point out that the use of URG for
PDU delimiting uses a service that is not present in the TCP
RFC, that services are part of the protocol specification, and
thus the proposed use is a change to the TCP RFC.

> Consequently you must remove this from the spec in the next
> version IMO...

Which is my opinion also.  But the call by the WG chair was for
concensus that allows the deployment of URG by allowing its
negotiation upon session login.

I'm now concerned with improving the hand-waving specifcation
of that negotiation into something solid.  Then people can
reliably defeat URG if implementation experience shows it
to be a Bad Idea (TM).  As I have no doubt it will :-)

Glen