[tsvwg] Re: editorial comments on draft-duke-tsvwg-udp-ecn-03: "ECN mark" vs "codepoint"

[Martin Duke <[e-mail]>]

I've filed Issue #15 to fix terminology.

Regarding Gorry's question, "did you test ability of a sender to set a
CE-Mark"

Not explicitly, no. I do have (Linux) unit tests that send CE marks to
localhost to verify that the reports are correct, and those seem to work.

If someone wants to test CE-sending, I will take the PR, but have no plans
to conduct this testing.

On Wed, Nov 13, 2024 at 11:13 AM Rodney W. Grimes <[e-mail]>
wrote:

> > On 13/11/2024 14:02, Neal Cardwell wrote:
> > > A quick editorial suggestion on?draft-duke-tsvwg-udp-ecn-03:
> > >
> > > It looks like this draft mostly uses variants of the phrase "ECN mark"
> > > (18 occurrences) to mean the two-bit ECN-related values in a packet,
> > > and only uses the phrase "codepoint" twice.
> > >
> > > RFC 3168 (classic ECN) is fairly consistent in using the phrase
> > > "codepoint" to refer to those two ECN-related bits. And RFC 3168 only
> > > uses the phrase "ECN marking" as the process of applying the CE code
> > > point ("replaced by CE when ECN marking occurs in?response to
> > > congestion or incipient congestion").
> > >
> > > RFC 9331 (L4S ECN) is in line with RFC 3168: the two bits are the
> > > "codepoint", "ECN mark" means CE, and "ECN marking" means changing the
> > > codepoint to CE.
> > >
> > > In line with RFC 3168 and RFC 9331, in my experience, the phrase "ECN
> > > mark"? is most commonly used for the CE code point.
> > >
> > > For clarity, I'd suggest that the?draft-duke-tsvwg-udp-ecn align with
> > > RFC 3168 and RFC 9331 terminology in this regard.
> > >
> > > best regards,
> > > neal
> > >
> > While we're discussing terms (and assuming you;re tracking issues...
> >
> > I'd personally prefer CE-Mark, rather than ECN-Mark, to avoid any
> ambiguity.
>
> Personally I would prefer the exactness of saying "apply the CE codepoint
> value".
>
> >
> > Another thing that comes to mind, is did you test ability of a sender to
> > set a CE-Mark, because some proposals have used that in the past as part
> > of their path probing to see if the receiver correctly reports the
> CE-Marks?
> >
> > I'd also encourage not using the word "TOS byte" if we can, because this
> > kind of perpetuates the idea that ToS exists, which can lead to some
> > akward pathologies with respect to DSCP and ECN codepoints, or I'd
> > suggest at least merits statements that ToS is not recommended any more.
>
> I agree on this point, TOS/ToS should not be used in any new documents,
> these fields are now known as the "DSCP codepoints and ECN codepoints".
> IMHO if TOS is mentioned it should be tightly coupled with the word
> "legacy", as in "the Legacy TOS byte."
>
> > Best wishes,
> > Gorry
> > (as an Individual)
>
> --
> Rod Grimes
> [e-mail]
>