[Top] [All Lists]

Re: [ietf-smtp] Good to see new list, comments about the "purpose"

2020-07-29 20:35:40

Some observations and an attempt to play devil's advocate:

(1) If we are going to have a separate list, let's have a
separate list, not create confusion by copying ietf-smtp and
having both the additional clutter and the risk of parallel
discussions.  So, while this note is copied to ietf-smtp for
obviously reasons, I hope it will be the last, or nearly-last,

(2) Perhaps I'm the only one, but I'd really like to try to
survive this week without putting something else on my virtual
plate.  So I would encourage others to just let this go until
next week (or longer).  In any event, for myself, please don't
expect further responses from me until at least the weekend.
And I would encourage not reading the rest of this note until
those who would do so have recovered from this week.

(3) Now the substantive part (and, for most of it, the devil's
advocate part):

Personally, I'm fine with the model of focusing on moving
5321bis and 5322bis to Internet Standard, using a separate {
Applicability Statement | BCP | Rubber Ducky } as a way to
accommodate whatever behavioral recommendations, comments on or
pointers to other specifications, and/or "roadmap" materials for
whatever mail-related specifications the WG concludes are
important.  As Pete said during the BOF about 5322[bis], I'd
really, really, like to get 5321[bis] off my plate, do so before
I retire [1], and do so with a minimum of pain and suffering.
The current plan appears to me to be the best way to accomplish
all of that.

Despite a BOF that was organized around that narrow scope and
focus, not only is it not the only way forward but it may be
based on a false premise.  RFC 6410 dropped the standards track
from three maturity levels to two in the hope that would result
in significantly more documents moving to Internet Standard.  A
handful of specifications that have been advanced without
changes to the underlying documents aside, it is rather clear
that has not happened and that, if anything, the Internet and
the users of IETF specifications care even less than they did
nine years ago.   I do care on grounds of tidiness; Pete,
Alexey, Seth, and Barry presumably care, and probably others
reading this care, but it is not clear that any significant
number of relevant others care.  And, while that narrow focus
may be the fastest way to get the work it defines done, it will
still be a fair amount of work; the BOF today provided evidence
that some things will be controversial and painful to resolve;
and so on.  The payoff may not be sufficient to justify that,
which suggests two other possibilities.

(a) RFC 5321 is a bad document (so was 2821).  It is really a
merge of three separate documents combined with excerpts from a
fourth and some added applicability statement material
associated with the technical specification.  In part because of
that merge and the age of RFCs 821 and 974, it switches
organizational styles and writing style from section to section.
The switch from the textbook BNF of RFC 821 to the ABNF of 2821
introduced errors; the transition to 5321 did not catch all of
them and may have introduced some new ones (see Appendix H,
Section H.1, of draft-klensin-rfc5321bis-03) and there is no
good reason to be confident that there are none left.  And, as
was sort of mentioned during the BOF, the references between the
two documents for syntax and the greater permissiveness of 5822
contribute to confusion.   Two consecutive WGs (DRUMS and YAM)
decided against doing a full rewrite --largely on the grounds
that it would take more time and risk more errors -- but, if we
really want to add value in this iteration, maybe we should be
developing something more like:

  (i) A complete rewrite of 5321
  (ii) An update to 5322 more or less along the lines planned --
material in 5322 only has one core document and two editors in
its history while I count about eight authors/editors for what
became 5321 -- but with some material torn out.
  (iii) A consensus, best practices, terminology and model
document using RFC 5598 as a starting point.
  (iv) A syntax specification that draws on both 5321 and 5322
and that is referenced by both.
  (v) A "roadmap" document for Internet email specifications,
possibly combined with A/S and/or BCP material for some of those
specs.  Or those could be one or more documents separate from
the roadmap.

Other breakdowns are clearly possible, but the goal would be to
recognize that we have something of a mess on our hands and that
part of the goal for producing Internet Standards is to make
them clear and easy to understand, a goal that actually requires
more than the narrow revisions now proposed.

(b) As any of us with experience in this area should have been
able to predict (and several of us did), mentioning the idea of
opening up 5321 and 5322 immediately drew multiple topics out
for consideration for inclusion in one or the other document (or
something related).  Maybe that energy is A Good Thing. If so,
perhaps we should put the efforts to get 5321bis / 5322bis to
Internet Standard aside for six months or a year (or longer) to
give those proposals time and attention sufficient to get clear
documentation and IETF consensus on whether or not they are good
ideas and then to be able to demonstrate interoperability, wide
deployment, etc., only then coming back to 5321bis/5322bis when
we are ready to make substantive (rather than procedural)
decisions about their inclusion.  That time might also allow
some already-published specs to mature sufficiently to be
incorporated or normatively referenced.

Again, I'm not suggesting either (a) or (b) above, but they
appear to me to be plausible alternatives that we should
consider and, if appropriate, reach consensus to explicitly
reject rather than blowing them off because they are
inconsistent with the path we have been on.


[1] Something that is feeling a lot closer right now than it did
last week.

ietf-smtp mailing list

<Prev in Thread] Current Thread [Next in Thread>