Re: [tcpm] Opsdir last call review of draft-ietf-tcpm-yang-tcp-06

["Scharf, Michael" <[e-mail]>]

Hi all,

To be more precise, we _have_ addressed parts of the OPSDIR review:

1/ We have described more precisely the differences to the TCP MIB, e.g., why the RTO configuration and the maximum connection limit is omitted. The RTO setup is simply only one of many algorithms of a modern TCP stack, and all have configuration parameters. And the RTO algorithm defined in the TCP MIB may not be the most relevant parameter in a modern stack.

2/ We have added the TCP connection state (i.e., the FSM) to the connection list. The description is aligned with the core TCP specification (793bis).

3/ We have also added the TCP listener list to better align with the TCP MIB.

Please have a look at the new version (https://datatracker.ietf.org/doc/html/draft-ietf-tcpm-yang-tcp-07) or the diff (https://www.ietf.org/rfcdiff?url2=draft-ietf-tcpm-yang-tcp-07) and let us know if more is needed.

The remaining suggestions in the review (e.g., TCP flags, congestion control algorithms, TCB control block, …) would be additions to the model beyond the TCP MIB. Also, that would significantly change the scope of the model. As already explained by Martin, there is no consensus in TCPM on such a model. A follow-up RFC could be published in that space, if there was enough energy and community consensus.

Best regards

Michael


From: Martin Duke <[e-mail]>
Sent: Thursday, June 16, 2022 11:00 PM
To: Gyan Mishra <[e-mail]>
Cc: [e-mail]; [e-mail]; Last Call <[e-mail]>; [e-mail] Extensions <[e-mail]>
Subject: Re: Opsdir last call review of draft-ietf-tcpm-yang-tcp-06

Gyan,

Thanks for your review. The authors have finally updated the draft, but mostly did not address your review.

It was explicitly not the intent of this effort to model every aspect of TCP implementations. There was strong WG consensus that this would be time-consuming and extremely unlikely to be deployed, as most TCP endpoints don't use YANG. Instead, this document is tightly focused on BGP routers, which do use YANG.

To include more aspects of TCP, we would want to see evidence that is relevant to this use case.

Martin Duke
Transport AD.

On Thu, Mar 3, 2022 at 12:29 PM Gyan Mishra via Datatracker <[e-mail]<mailto:[e-mail]>> wrote:
Reviewer: Gyan Mishra
Review result: Not Ready

The TCP FSM is of course the most critical component of the transport.

It would be good to list all that is not included in the Yang model that exists
in the TCP MIB.  Also would be helpful as to reasons why.

I see mentioned that RTO is not part of the Yang model, however to be complete
I think it should be included.

The yang model seems to not have  all the TCP FSM states listed below:
Closed
Listen
SYN RCVD
SYN Sent
Established
FIN wait 1
FIN wait 2
Closing
Time wait
Close wait
Last Ack

Also the Yang model does not reference the TCP Flag bits set during state
changes in the FSM below as well as flag combinations for example for
establishment state you sent SYN, receive SYN/ACK, ACK

URG
ACK
PUSH
RESET
SYN
FIN

I also don’t see anything in the Yang model on TCP window and window scaling
and CWIN congestion control algorithm backoff.

Also I don’t see any mention in the Yang model about the well known port range
0-1023 and > 1023 anonymous port range for the TCP socket to be established.

Also mention about the TCP TCB control block.

Local IP Local Port
Remote IP Remote Port
Interface
Process
State
Local/Send window
Remote/Receive window
Send SQ Ack
Send SQ Un-ack
Send SQ Next
Not to be sent
Receive Next
RTT
Buffer pointer