Hi Greg,

draft-zheng-mpls-ls-ping-yang-cfg defines transmit interval in RPC because all 
ping operations are done via RPC.  I do not consider BFD echo to be "on demand" 
like LSP Ping (caveat: this is possibly due to the BFD 
configuration/implementation I am most familiar with).

Regards,
Reshad.

From: Greg Mirsky <[email protected]<mailto:[email protected]>>
Date: Monday, February 27, 2017 at 1:07 AM
To: Mahesh Jethanandani 
<[email protected]<mailto:[email protected]>>
Cc: Reshad <[email protected]<mailto:[email protected]>>, 
"[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>, 
"[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>
Subject: Re: Correcting BFD Echo model

Hi Mahesh,
BFD Echo transmit interval is not part of RFC 5880, only Rx interval is. And Rx 
interval is sufficent to reflect whether local system is willing to receive BFD 
Echo messages from the particular BFD peer. Introduced 
desired-min-echo-tx-interval functionally overlaps with the standard-defined 
required-min-echo-rx-interval. Hence my suggestion to remove 
desired-min-echo-tx-interval from grouping bfd-grouping-echo-cfg-parms. But 
operators need a way to specify transmit interval for on-demand OAM command 
like BFD Echo, IP ping or LSP ping. I couldn't find YANG model proposal for IP 
ping but in draft-zheng-mpls-lsp-ping-yang-cfg transmit interval used in RPC, 
not as part of configuration.

Regards,
Greg


On Sun, Feb 26, 2017 at 8:33 PM, Mahesh Jethanandani 
<[email protected]<mailto:[email protected]>> wrote:

On Feb 26, 2017, at 4:39 PM, Greg Mirsky 
<[email protected]<mailto:[email protected]>> wrote:

Hi Reshad,
thank you for the question. Here's my reasoning:

  *   only Required Min Echo RX Interval is present in RFC 5880 and it allows 
to indicate not only the smallest interval between consecutive BFD Echo packets 
but whether system supports BFD Echo function at all;
  *   since BFD Echo may be transmitted only when the session state is Up, 
operator is fully equipped to learn the value of Required Min Echo RX Interval 
of its BFD peer and to set Echo transmit interval accordingly;
  *   requesting BFD Echo, in my opinion, is no different from requesting IP 
ping or LSP ping.

Hence my conclusion - transmit interval for BFD Echo is more suitable in RPC 
then as configuration parameter.

I do not think that is reason enough for it to be a RPC.

A RPC is an operation one defines in the YANG model specifying both input and 
output parameters. There are no operations to be had here.

And the definition and desired behavior of desired-min-echo-tx-interval is not 
very different from required-min-echo-x-interval. It is as the definition says, 
a configuration parameter that can be set, with zero having a special meaning 
in both cases.


Regards,
Greg

On Sun, Feb 26, 2017 at 2:52 PM, Reshad Rahman (rrahman) 
<[email protected]<mailto:[email protected]>> wrote:
Hi Greg,

Can you please explain why you believe this should go in RPC?

Regards,
Reshad.

From: Greg Mirsky <[email protected]<mailto:[email protected]>>
Date: Saturday, February 25, 2017 at 6:48 PM
To: "[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>, 
"[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>
Subject: Correcting BFD Echo model
Resent-From: <[email protected]<mailto:[email protected]>>
Resent-To: <[email protected]<mailto:[email protected]>>, Reshad 
<[email protected]<mailto:[email protected]>>, 
<[email protected]<mailto:[email protected]>>, 
<[email protected]<mailto:[email protected]>>, 
<[email protected]<mailto:[email protected]>>
Resent-Date: Saturday, February 25, 2017 at 6:48 PM

Dear All,
I've reviewed the BFD YANG model and now I'm thinking that 
desired-min-echo-tx-interval and attributing to it the behavior, i.e. when the 
value is 0, of Required Min Echo RX Interval are not in the right place. I 
think that definition of desired transmit interval of BFD Echo should be in 
corresponding RPC definition, not in configuration part of the model.
Appreciate your comments.

Regards,
Greg



Reply via email to