Re: [tsvwg] L4S Review

[Bob Briscoe <[e-mail]>]

Nico,

On 03/12/2018 09:28, Kuhn Nicolas wrote:
>   > Also, another network component on the path could exploit this  > semantic for other purpose.
>
> [RS] There have been measurements in the public internet, which show no use of ECT(1) at all. Also, as these drafts are experimental, private environments could choose not to participate.
>
> However, if you know of a concrete example (SatCom?) where ECT(1) is used as a specific signal (possibly between network elements, bleaching out again before the end systems can observe it), please speak up!
>
> [NK] I am not aware of any example in that sense. That being said, TCP-splitters may break the end-to-end signaling.
[BB] Back-to-back full TCP stacks in the middle would be fine with L4S. 
Whether they both supported L4S or not, they would each work correctly 
for their leg of the e2e connection, assuming they negotiated ECN 
feedback to the specs. If these middle stacks supported AccECN and L4S 
then, even if only one of the 'real' ends supported these, that segment 
of the connection could use L4S, even if the other segment didn't. That 
could be a very useful deployment enabler.

However, getting back to your point with Richard, not understanding and 
therefore breaking ECN semantics isn't really an example of exploiting 
the semantics for another purpose. As in my previous email, can you give 
an example of what you were trying to mean?


Bob

PS. Again, sry for delayed response - hope you haven't completely 
forgotten the context.

-- 
________________________________________________________________
Bob Briscoe                               http://bobbriscoe.net/