* Section 22: Specify that RES may be sent other than on return traffic only when DLPMTUD is enabled.

[Mike-Heard]

Raised by Erik Auerswald. See https://mailarchive.ietf.org/arch/msg/tsvwg/qlcFuZjzlzxyoFg2UOhv-PK8Kag/

[auerswal]

Suggested text for an adjusted REQ/RES description in https://mailarchive.ietf.org/arch/msg/tsvwg/mdPPObp-ULgSWWQnJbh3SEeYdTE/

[gorryfair]

I'm OK with the proposed text, as long as we explain the application isn't always the "application", but it is always a protocol above UDP Options.

[auerswal]

Yes, the "application" can be an actual application, a library, or part of the operating system, to give a few examples. Any "user" of UDP Options would be an "application." Perhaps extend "provide a means for applications" to say "provide a means for applications (i.e., any users of UDP Options)"?

[jtouch]

I updated the text in -23 to try to explain this. Note the security caution in that section.

[gorryfair]

The new words look good, but we ought to have gone further when we note DPLPMTUD …. initiate UDP responses in the absence of user transmission

• I think we ought to go further and add words that clarify that the RES response is only when the receiver has enabled for DPLPMTUD for the socket pair.

[jtouch]

REQ/RES is not designed solely for DPLMTUD. RES happens only when DPMPLUD (or some other middle layer/OS) sends it and the default is to NOT send it; the UDP layer NEVER sends RES of its own accord.

[Mike-Heard]

I have to agree with Joe on this point ... the UDP options spec should not mandate that DLPMTUD is the only allowed use of REQ/RES.

[gorryfair]

We are close to agreeing, but the thread is confusing. I was wanting an example, as in the DPLPMTUD spec for UDP Options that adds that The new words look good, but we ought to have gone further when we note DPLPMTUD …. initiate UDP responses in the absence of user transmission

I think we ought to go further and add words... this is what the spec says for the way DPLPMTUD will use this:

"Use of DPLPMTUD MUST be explicitly enabled by the application, for instance once an application has established connectivity and is ready to exchange data with the remote Upper Layer protocol. Similarly, a receiver SHOULD NOT respond to a REQ Option until DPLPMTUD has been enabled."

[gorryfair]

I should have been clearer, the sentence I wished to suggest was: "For example, an application has to explicitly enable the generation of a RES response by DPLPMTUD when using UDP Options [ID.REF]."

[jtouch]

Text added as Gorry suggested above in -25. Please let me know if that closes this issue or if additional text is needed.