Re: [tsvwg] RDMA Support by UDP FRAG Option

[Joseph Touch <[e-mail]>]

Hi, Mike,

> On Jun 21, 2021, at 12:48 PM, C. M. Heard <[e-mail]> wrote:
> 
> On Mon, Jun 21, 2021 at 11:19 AM Joe Touch <[e-mail] <mailto:[e-mail]>> wrote:
>> On Jun 21, 2021, at 9:37 AM, C. M. Heard <[e-mail] <mailto:[e-mail]>> wrote:
>> On Mon, Jun 21, 2021 at 8:48 AM Joseph Touch wrote:
>> …
>> > In my opinion, the best chance for UDP options to be deployable and
>> > successful is to require that the surplus area always sums to zero.
>> 
>> When UDP CS==0, OCS provides that capability.
>> 
>> It would, if the requirement for a receiver to check OCS when it is present were dropped when UCP CS==0.
> 
> Why / how does this help? 
> 
> You said (or at least I thought you said) that when UDP CS==0, OCS provides the capability to force the surplus area to sum to zero.

OCS works like CCO did mostly, at least computationally - or at lest that’s what it’s supposed to do.

> If I understand the current spec correctly, then that is not really true.
> 
> It's true that OCS is optional when UDP CS==0, but my understanding is that per the current spec, if it is present, the receiver is required to check it and drop the options if it does not follow the rule specified for OCS.

Yes, and that makes sense - if you’ve added OCS, regardless of whether the UDP CS==0 or not, then you want the options checked. All we currently have been saying is that “OCS can be omitted when CS==0”, not that it MUST be.

> And that rule is not that the surplus area sums to zero, but that the surplus area plus the pseudo-header consisting of the surplus are length sums to zero.

It was originally (through draft-ietf-tsvwg-udp-options-05)  to cover the surplus area only; it was in the additional CCO-like behavior that it changed to add the pseudo header  (the surplus length). So are we going back to the pre-CCO definition? 

Joe