Rare project developers

[mc36 <>]

nevermind, if the draft/rfc/nexus captures indicate the possibility of header 
_insertion_ then we'll have int, it's on the todo already...
and we'll be able to measure how precise it is in java or in the tofino.... 
:)))))
thanks for the research!
regards,
cs


On 4/8/21 7:26 PM, Ronald van der Pol wrote:
> On Thu, Apr 08, 2021 at 14:58:24 +0000, Tim Chown wrote:
>
> > One of the open issues is with the clock synchronisation, as reported above.
>
> Hmm. page 27:
> "Another constraint to consider when creating P4 programs is that
> Tofino chips do allow complex computations on headers to maintain line
> rate, such as multiplication to convert the time expressed in nanoseconds
> to other time units (e.g., microseconds)."
>
> I think "... do NOT allow ..." was intended.
>
> And I am not sure I agree. I think the hash unit can shift and
> thus multiply by 1024. But it does not solve the wrap-around.
> I don't understand the reasoning.
>
> But maybe I don't understand what this paragraph intents to say.
>
> > I guess most NRENs will likely see IAOM (implemented by Cisco, Juniper, etc)
> > than INT, but to have an INT implementation that NRENs can experiment with
> > would be interesting.
>
> I agree with Csaba here. If we implement it, why not make it
> such that it operates with Cisco/Juniper? The differences are
> not that large.
>
>      rvdp