On 27.6.2015 17.55, Juliusz Chroboczek wrote:
Markus,
Section 4.4 of -06 seems to say that Node Endpoint TLVs MUST be silently
discarded.
Not really, it is not unknown TLV as supporting it is required. However,
wording deltas accepted if it makes it clearer.
Section 5, "containing data about the most recently sent RNS", should
that be per interface ("endpoint")? Since there's a distinct trickle
instance per interface, this would make sense to me.
It describes a minimal implementation (and that global counter covers
the minimum needed RNS sending rate limiting). A possibly more 'correct'
one would be e.g. tied to Trickle instances, which may be either per
endpoint, or per (peer, local endpoint) in unicast case.
Since there's state shared between the Endpoint TLV and the other TLVs
in a datagram, shouldn't you be requiring that the Endpoint TLV be sent
first?
We used to have a rather strict ordering of message contents (e.g. same
as inside node data). It was not found desirable by reviewers. As a
result of that, currently we do not force _any_ ordering.
I think you have a valid point here though; as the (datagram-case)
packet processing can be made single pass given ordering constraint, it
does make sense.
What do you think would be best here? Just specify that it MUST be
before a Network State TLV within datagram? Or something more broad,
like first TLV? (possibly too broad for future?)
Cheers,
-Markus
_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet
