Re: [mpls] draft-fang-mpls-label-forwarding-no-swap - how much does it really save?

[Andrew Qu <[e-mail]>]

Hi Andy,

I believe it is RFC author’s duty to make RFC as clear as possible so that reader/implementation engineer to be clear
About the intent of authors.

In addition to what I brought up in my previous mail about the huge saving in HW resources when if on paper we have
Additional “no-swap” operation added to existing MPLS operation, I believe we should define “no-swap” first,
Let implementation folks to decide whether use “swap” operation logic to achieve “no-swap” or use specific
Way of achieving that.

Why RFC authors dictate the “real implementation” ?  all we need is just the “final compliant” result, however
“no-swap” is simply different than “swap”, why we ignore the fundamental difference.

Thanks,

Andrew

From: Andrew G. Malis [mailto:[e-mail]]
Sent: Thursday, July 23, 2015 10:33 PM
To: Shahram Davari
Cc: Loa Andersson; Andrew Qu; Alexander Vainshtein; Stewart Bryant; S. Davari; [e-mail]
Subject: Re: [mpls] draft-fang-mpls-label-forwarding-no-swap - how much does it really save?

Sharam,

The point was made during the meeting that on the wire, "no swap" (but still decrement TTL, etc.) is indistinguishable from "swap to the same label". So this really is a local optimization in the switch, and not subject to standardization.

Cheers,
Andy

On Thu, Jul 23, 2015 at 7:18 PM, Shahram Davari <[e-mail]<mailto:[e-mail]>> wrote:
I thought the idea was to decrement TTL and consider EXP bits for queuing and count as a hop so you can do OAM, etc.. Basically swap to itself.

Thx
Shahram