--On Thursday, May 05, 2011 09:13 -0700 The IESG
<iesg-secretary(_at_)ietf(_dot_)org> wrote:
The IESG has received a request from an individual submitter
to consider the following document:
- 'Reducing the Standards Track to Two Maturity Levels'
<draft-housley-two-maturity-levels-06.txt> as a BCP
The IESG plans to make a decision in the next few weeks, and
solicits final comments on this action. Please send
substantive comments to the ietf(_at_)ietf(_dot_)org mailing lists by
2011-06-02.
...
Hi.
I made several comments about earlier versions of this document
then stopped because, while the incremental versions have all
included small improvements, they have addressed few of the
basic issues and the document seemed to be riding something of
a juggernaut. Because it is now in Last Call, I want to
summarize and update those comments, in large measure because I
don't want my near-silence since late January to be taken as
agreement.
This proposal is problematic in several ways. The first is
obviously the most important, but the others suggest changes
that might be useful whether or not that main provision is
adopted.
(1) Excessively-high requirements for Proposed Standard and
dropping the three-level model.
The document asserts, I believe correctly, that one of the
problems with the IETF Standards Process is that the
relatively low bar established for Proposed Standards in
RFC 2026 has evolved into a very high bar and that we
should return to that lower requirement level. Yet it
effectively proposed to solve that problem by reducing the
number of steps in the standards process to two. No
evidence has been offered that whether or not the standards
process is two levels or three has anything to do with
difficulties completing Proposed Standards or getting them
to a satisfactory level of completeness and correctness.
The logic on that subject in the document appears to me to
quite similar to a physician saying "We agree there is a
problem with your hand. We don't know how to treat it
either surgically or medically, but we should do something,
so let's amputate your foot". That really makes no sense
and is probably bad for the patient (only probably... one
really never knows).
In addition, draft-housley-two-maturity-levels-06 does not
discuss any mechanism for getting from the current
as-practiced requirements for Proposed Standard back to the
requirements of RFC 2026, a schedule for doing so, nor
whether it is necessary to warn readers about the level of
scrutiny applied (it may not be, but there has been little
or no discussion of that subject in the IETF and there is
no such discussion in the document). If this were a
technical specification, unless there were an expectation
that a transition will occur immediately and by magic, that
omission would be a "known defect".
Suggestion: The level of completeness and quality required
for a Proposed Standard, at least insofar as those
requirements exceed the requirements in RFC 2026, are
entirely in the hands of the IESG, relevant WGs, and the
Last Call process. Whether those requirements can be
reduced to the level required by 2026 has nothing to do
with this document because there is no change to the 2026
requirements. If we are serious about making that change,
I recommend that the IESG:
(i) Put this document aside temporarily
(ii) Announce to the community (presumably via an IESG
statement) that, as a general matter, documents
submitted for Proposed Standard will not be
required to meet conditions beyond those imposed
by 2026. Note that 2026 allows imposition of
additional requirements, especially requirements
for implementation experience. Whether a
particular WG should strive for a higher quality
standard for its documents should be treated as a
management issue (e.g., it would rarely make sense
to downgrade documents that were nearly finished
at a higher level), but reviews on Last Call or
from within the IESG that suggest a higher level
of perfection should probably be ignored unless
they provide reasoning for the greater requirement.
(iii) The above implies that different areas, and
perhaps even different WGs within an area, will end
up with different practical requirements based, at
least in part, on their perception of the risks
implied by a less comprehensive document. I see
that as an advantage and recognition of reality,
not a problem.
(iv) After a couple of years -- probably the minimum
needed to achieve a steady state given documents
already under development and targeting a greater
level of requirements than 2026 anticipated -- the
IESG should review the situation and make a
recommendation to the community as to whether
further changes are needed to the standards process
based on experience with this revised way of
dealing with proposed standards.
Note that the model described above is different from the
one in draft-bradner-restore-proposed-00, especially with
regard to schedule. draft-bradner-restore-proposed-00
suggests a fixed transition schedule; these notes propose
to make transition a management matter (and, indirectly, a
test of the IESG's seriousness about restoring the intended
Proposed Standard definition); as noted above,
draft-housley-two-maturity-levels-06 does not address this
issue. After a few months of consideration, I now believe
that "management matter" approach provides both a better
transition plan and long-term framework than trying to
force documents into a single level of completeness and
scrutiny. The other comments and suggestions in
draft-bradner-restore-proposed-00 could be considered
useful supplements to the suggestions above.
(2) Elimination of the "downref" requirement for
standards-track documents.
The original motivation for the "no normative references to
documents at lower maturity levels" rule was to be sure
that the documents referenced were of adequate quality to
be used, effectively, as part of the document under
consideration. If one thinks about it, having a document
that the IETF certifies as representing an implemented,
deployed, and useful specification dependent on something
that is no more that an untested probably-good idea makes
little or no sense. Referencing Proposed Standards that
have survived the level of scrutiny that has prevailed in
the last several years should not be nearly as
problematic, but removing the downreferencing requirement
at precisely the same time that we intend to get Proposed
Standards back to the 2026 level creates exactly the risk
that the rule was intended to prevent.
Suggestion: The community had this almost correct in RFC
4897 in that the community should be able to review
proposed downward references. RFC 4897 more or less
assumes that the review should require explicit community
review and signoff. That might be excessive; a more or
less default review might be sufficient. But we should
preserve the community's ability to conclude that a
downreferenced document is inadequate for the purposes of
the referencing document, if only through a Last Call
response. The change made in
draft-housley-two-maturity-levels-06 appears to eliminate
"the referenced specification is of inadequate quality to
be used as part of this specification in a normative
reference". It seems to me that is a bad idea and probably
even an unintended side-effect.
Note that this suggestion should be as effective at
eliminating downward references as a "major cause of
stagnation in the advancement of documents" as the proposal
in draft-housley-two-maturity-levels-06 if, in fact, that
is a major issue.
(3) Maturity level, document editorial quality, and RFC Editor
issues
The rather cavalier assumption that a three-stage standards
process somehow makes it less possible to apply the
Proposed Standard criterion of RFC 2026 omits another
important reason why the bar for Proposed Standards has
become so high. It would be a reasonable (and arguably
historically accurate) statement of the traditional
editorial quality requirement for Proposed Standard
documents is that they needed to be good enough to be used
for implementation among cooperating parties, i.e., clear
enough that an implementer with access to WG mailing lists
(or authors) could implement them and evaluate document
quality for implementation. Perfect English writing style
was not a requirement, nor was sufficient clarity that one
was assured of being able to do an interoperable
implementation in a "clean-room" environment relative to
the authors and discussions about development of the spec.
Viewed from the editorial quality perspective, the Draft
Standard level was the one at which document quality was
expected to be improved to meet those much higher standards
of editorial quality and technical completeness and
clarity. RFC 2026 requires "no known omissions" for
Proposed Standard, not "no omissions" or strong assertions
of completeness.
One of the significant obstructions to getting Proposed
Standard documents published quickly is that the IESG has
insisted on a rather high level of technical quality,
usually doing the editing itself rather than trusting the
RFC Editor staff to get that right. WGs and the community
have responded to the expectation of IESG insisting on
documents of very high editorial quality (and sometimes the
expectations that documents will never advance beyond
Proposed Standard) by nit-picking editorial quality
internally and on IETF Last Call, creating additional
delays. If we are going to get Proposed Standards
published more quickly and under 2026 criteria, the IESG
will need to do some self-examination about these editorial
document quality issues and perhaps either settle for lower
quality at Proposed, shift more of the editorial burden to
the RFC Editor, or both. In any case, if documents are to
be less than editorially perfect at the Proposed Standard
level, there needs to be provision, or at least an
understanding, for assuring that documents meet a higher
standard at subsequent maturity levels. Those provisions
are absent from draft-housley-two-maturity-levels-06, as is
even a statement of the possibility that they might be
relevant... another "known omission".
Note that accepting a less formal, and less
formally-correct, editorial style is consistent with
publishing "Proposed Standards as soon as rough consensus
is achieved". Spending weeks or months sorting out
editorial issues is not.
In addition to the above, Long-duration MISSREF entries in
the RFC Editor queue are a major source of the claim that
downward referencing requirements are a significant
impediment to advancement of documents. However, if those
entries are examined, a significant fraction of them are
actually either missing references pointing to I-Ds (for
which the present proposal provides no help at all) or a
mechanism for identifying a set of documents that have to
be published together to make sense. As long as we
periodically approve and publish a single logical standard
as a collection of documents that may go through a WG or
the IESG at different rates, there will be a need to hold
the first-approved of those documents until the
last-approved one is ready.
draft-housley-two-maturity-levels-06 does not make
provision for whatever additional RFC Editor states are
needed in the IETF Stream to cover that problem. While not
major, it is another sign that, if the present proposal
were being evaluated as a technical specification, it would
be found to contain known omissions.
(4) Discussion of the STD document series in Section 6.
It has been pointed out several times, in recent months to
suppress discussion of proposals that might be alternates
to some or all of this one, that there are many issues with
how the IETF develops, approves, documents, and manages
standards that are not addressed in this specification and
that the community cannot reasonable expect to address all
at the same time. The question of what should be done
about the STD numbers is just one entry on this very long
list. Examination of other IETF procedural RFCs indicates
that we rarely, if ever, include a discussion of a single
option that we have decided to _not_ pursue (even if only
in the near term). The only apparent justification for
including Section 6 in this draft is that there was a
specific proposal in an earlier version, a proposal that,
as the draft indicates, did not get community consensus.
The section should be dropped from this document before
publication. If the community wants to have a series
discussion of STD numbers, I suggest that
draft-klensin-std-numbers might be one contribution to that
discussion.
(5) Patents
The specification says (last bullet in Section 2):
> * If patented or otherwise controlled technology is
> required for implementation, the separate
> implementations must also have resulted from separate
> exercise of the licensing process.
While this text is copied from 2026, things have changed
and the requirement has been interpreted very loosely,
particularly to leave interpretation of "required" in the
hands of the implementers.
If it is repeated in a spec that is approved now and that
claims to reset and clarify Draft Standard requirements, it
can easily be read as a requirement that the community has
to somehow evaluate whether claimed controlled technology
is actually "required for implementation". That may be
obvious in some cases; it may be highly controversial and
dependent of validity of patent claims (something that can
normally be resolved only by the courts) in others. The
IETF has been repeatedly advised, both by legal counsel and
by various incarnations of IPR WGs, to avoid straying into
the territory in which it takes a position on whether an
IPR claim is valid. Consider a scenario in which some
company, say ChuckCo, decides that an IETF Internet
Standard for Alice's proposal would put them at a
competitive disadvantage. It is merely necessary for
ChuckCo to file an IPR claim against the technology in
Alice's spec, even a obviously spurious one, assert a
completely irrational and punitive licensing model that
cannot be exercised in a reasonable way, and then sit back
and watch the IETF paralyze itself over this provision as
written.
Suggestion: I believe that advice of Counsel should be
sought about this language and, if feasible, alternate
language produced that either turns the requirement into an
"if feasible" request or that places the responsibility for
evaluating whether the claims are applicable on the
implementers and not on the IETF or IESG.
Conclusion:
This document is not ready for approval. It is justified
in terms of one issue but offers no evidence that it will
help at all with that issue and instead proposes a solution
that is associated with another set of issue that may not
be critical-path. For that primary issue (speeding
documents into Proposed Stanard), it fails to address known
issues and, more important, is not required at all. While
"no known omissions" is not formally a requirement for IETF
procedural documents to be published in the BCP series, it
is nonetheless a useful guideline and this document does
contain such omissions in several areas. It removes an
important criterion for evaluating document readiness for
publication when the document depends on other
specifications that are unclear or worse. It makes no
provision for transition. It contains spurious and
almost-irrelevant text. And it potentially opens an IPR
can of worms that the IETF has been warned against.
thanks for listening
john
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf