One nice thing about CoAP is that it has also been defined/used over other
transports, that are not UDP. Like directly over a serial link or over SMS.
This was another potential use case: to use GRASP when there's not even IP
connectivity yet.
The charter says this about a potential new "lightweight GRASP":
Generic use cases of Autonomic Network and new GRASP extensions/options for
them
If one sees it as an extension of GRASP (ability to run over non-TCP, UDP, or
non-IP transports) then it does fit in charter.
Defining some use cases would be relevant also - if it's a wired ANI then we
don't have a huge amount of changes in links over times. If it's a wireless
(e.g. 6LoWPAN) ANI then this may be the case.
Even if mesh links are highly variable over time, resulting in more packet loss
(e.g. 6LoWPAN situations), TCP may still work ok if it's tuned to this use
case. OpenThread's TCP implementation was I believe tuned in this way based on
work of Sam Kumar
(see https://www.usenix.org/conference/nsdi20/presentation/kumar ) . In
Thread's case, even when links change and mesh topology changes the nodes still
keep their original IPv6 address so the TCP communication can continue in
principle. I never tested TCP performance yet in this case, but it would be
interesting to see if there's a problem there or not, depending on the level of
links change.
Esko
-----Original Message-----
From: Anima <[email protected]> On Behalf Of Brian E Carpenter
Sent: Monday, March 18, 2024 09:38
To: 蒋胜 (Sheng Jiang) <[email protected]>; Anima WG <[email protected]>
Subject: [Anima] Lightweight GRASP
I think it would be good to see a first draft on this, with some ideas about
UDP and COAP instead of TCP.
Regards
Brian
_______________________________________________
Anima mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/anima
_______________________________________________
Anima mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/anima
