Re: [tcpm] rto-consider: "regularly" taking RTT samples

[G Fairhurst <[e-mail]>]

Hi Mark!

On 22/11/2019, 21:45, Mark Allman wrote:
> Hi Gorry!
>
> Thanks for all the comments.  I am going to work through them and
> hack on the document.  I am addressing a couple of the easier
> comments today.  But, will address them all soon.
>
> I-D>>  In steady state the RTO SHOULD be set based on recent
> I-D>>  observations of both the FT and the variance of the FT.
>
> Gorry>  - I don’t think these are wrong, but I think the advice is
> Gorry>    not very crisp for a BCP - how recent? How does this relate
> Gorry>    to Comment 3?
>
> The "recent" here is defined in requirement (2b), as you note.  So,
> it isn't meant to be crisp here.  I can see two paths:
>
> (1) Remove the word "recent" from here.  The overall point of
>      guideline (2a) is really that the RTO SHOULD be based on
>      observations---i.e., not statically set.  The notion of "recent"
>      isn't hugely important here as it is concretely developed in
>      (2b).
>
> OR,
>
> (2) We could leave the "recent" in (2a) with a forward reference to
>      indicate we are not leaving this as nebulous somehow.  E.g.,
>      "... recent observations---as defined below---of both ..." if
>      folks think keeping this notion (colloquially) in (2a) is a good
>      idea.
>
> I am pretty indifferent to both options.
Omitting the word (a) is indeed easy. I suspect the new formulation of 
(b) effectively defines regularly, so a recent sample should be available.
> I-D>>  FT observations SHOULD be taken regularly ...
> I-D>>
> I-D>>  The notion of "regularly" SHOULD be defined as at least once
> I-D>>  per RTT or as frequently as data is exchanged in cases where
> I-D>>  that happens less frequently than once per RTT.
>
> Gorry>  - To me, reads as impossible RFC2119 text, is this a
> Gorry>    definition or something else? or Designs are REQUIRED to
> Gorry>    define ... etc.
>
> I really can't parse the complaint here (which is probably because
> you well explained it, but since I wasn't there the slide isn't
> conveying that to me).  I don't see this impossibility you are
> suggesting.
Sorry, happy to explain why I thiunk this can be misread, but reading-on 
we already have a solution that works for me.
> That said,
>
> Gorry>  I hope this actually is intended to be:
> Gorry>  FT observations SHOULD be taken be at least once
> Gorry>  per RTT or as frequently as data is exchanged in cases where
> Gorry>  that happens less frequently than once per RTT.
>
> This is in fact the intent and I am happy to use those words.
Then let's do that.
> allman
Gorry