[clue] Benjamin Kaduk's Discuss on draft-ietf-clue-protocol-17: (with DISCUSS and COMMENT)

[Benjamin Kaduk <[e-mail]>]

Benjamin Kaduk has entered the following ballot position for
draft-ietf-clue-protocol-17: Discuss

When responding, please keep the subject line intact and reply to all
email addresses included in the To and CC lines. (Feel free to cut this
introductory paragraph, however.)


Please refer to https://www.ietf.org/iesg/statement/discuss-criteria.html
for more information about IESG DISCUSS and COMMENT positions.


The document, along with other ballot positions, can be found here:
https://datatracker.ietf.org/doc/draft-ietf-clue-protocol/



----------------------------------------------------------------------
DISCUSS:
----------------------------------------------------------------------

Thanks for the generally clear and well-written document!
I would like to discuss whether there needs to be more prominent coverage
of timers/timeouts, especially as relating to the state machines.  (I'd be happy
to learn that this is well-covered elsewhere in the document set; I just haven't
run into it yet.)

In a similar vein, do we want to have any treatment of avoiding infinite loops
(e.g., when a 'configure' or 'advertisement' is rejected in expectation of modification
but the sending implementation continues to generate an identical message)?

It is not clear to me that any change to the document text is needed in either case,
but I don't know to what extent the topics have already been discussed.


----------------------------------------------------------------------
COMMENT:
----------------------------------------------------------------------

I also have some substantial comments that do not rise to Discuss-level.

How do I know which endpoint is the channel initiator and which is the
channel receiver?
draft-ietf-clue-signaling suggests that the DTLS client is the channel
initiator, but even that is not explicit about it -- the protocol could be
considered under-specified if there is insufficient clarity.

Section 2

The MCU definition doesn't actually expand the acronym, which seems a
little reader-unfriendly.

Section 5.1

There are perhaps more XML extension points in here than is reasonable for
some of these elements (e.g., <versionsListType>).

Section 5.2

                           If the responseCode is between 200 and 299
   inclusive, the response MUST also include <mediaProvider>,
   <mediaConsumer>, <version> and <commonExtensions> elements;

Maybe re-mention that MP and MC are booleans here.

   Finally, the commonly supported extensions are copied in the
   <commonExtensions> field.

Does this need to say that only extensions that are applicable to the
negotiated protocol version are included?  (Also, how does one handle an
extension that exists for multiple major versions -- are there two
<extension> elments for it in the <options> message?)

   Upon reception of the 'optionsResponse' the version to be used is the
   one provided in the <version> tag of the message.  The following CLUE
   messages MUST use such a version number in the "v" attribute.  The
   allowed extensions in the CLUE dialogue are those indicated in the
   <commonExtensions> of the 'optionsResponse' message.

Couldn't this restriction on the "v" value apply even to the
'optionsResponse' message?

Section 5.3

   The 'advertisement' message is used by the MP to advertise the
   available media captures and related information to the MC.  [...]

I'd consider avoiding the definite article "the" to refer to MP/MC roles,
since in many caess there will be 2+ of each, and we don't want to confuse
the reader into thinking that there is an MP/CR equivalence or something
like that.  So, perhaps "each MP" and "the corresponding MC".

Section 5.4

                             As it can be seen from the message schema
   provided in the following excerpt (Figure 6), the 'ack' contains a
   response code and a reason string for describing the processing
   result of the 'advertisement'.  [...]

[the reason string is part of the base clueResponseType]
The text quoted here could be read as implying that the reason string is
required in the 'ack' message, a stronger requirement than of the base
clueResponseType where it has minOccurs=0.  Some greater clarity in the
text here is probably called for, especially since when the 'ack' is
piggybacked on a 'configure' message, there is no provision for a reason
string at all.

Section 5.5

                             The <ack> element MUST NOT be present if an
   'ack' message has been already sent back to the MP.

I think you need to clarify that this is scoped to the current
advSequenceNr.

Section 5.6

                                             It contains (Figure 8) a
   response code with a reason string indicating either the success or
   the failure (along with failure details) of a 'configure' request
   processing.  [...]

[Same comment about reason string as for 'ack']

Section 5.7

   Such new response codes MUST NOT overwrite the ones here defined and
   they MUST respect the semantics of the first code digit.

nit: is this "overwrite" or "override"?

   This document does not define response codes starting with "1", and
   such response codes are not allowed to appear in major version 1 of
   the CLUE protocol.  The range from 100 to 199 inclusive is reserved
   for future major versions of the protocol to define response codes
   for delayed or incomplete operations if necessary.  Response codes
   starting with "5" through "9" are reserved for future major versions
   of the protocol to define new classes of response, and are not
   allowed in major version 1 of the CLUE protocol.  Response codes
   starting with "0" are not allowed.

This text seems to also preclude extensions to major version '1' from
defining 1xx or [5-9]xx reason codes; is that the intent?

Section 6

   When the CLUE data channel set up starts ("start channel"), the CP
   moves from the IDLE state to the CHANNEL SETUP state.

nit: only one of "sets up" and "starts" is needed.

   When in the ACTIVE state, the CP starts the envisioned sub-state
   machines (i.e., the MP state machine and the MC state machine)
   according to the roles it plays in the telepresence sessions.  Such
   roles have been previously declared in the 'options' and
   'optionsResponse' messages involved in the initiation phase (see
   OPTIONS sections Section 5.1 and Section 5.2 for the details).  [...]

My reading of the initiation phase is that each CP sends only a boolean
indication of whether it supports the MP/MC roles, and so each party has to
determine on its own whether it will act as a MP and/or MC; is that
correct?  If so, do we need to say anything about how the boolean matrix
translates to activating the respective sub-state machines?

Section 6.1

                         'configure+ack' messages referring to out-of-
   date (i.e., having a sequence number equal to or less than the
   highest seen so far) advertisements MUST be ignored, i.e., they do
   not trigger any state transition.  [...]

Is this really less than or equal or just less than?  Also, is "seen" the
right verb, since IIUC these are sequence numbers that the MP has
*generated* in its advertisements?

Section 7

                                                          In other
   words, in this example, the MP MUST use version 1.4 and downgrade to
   the lower version.  [...]

nit: does the phrase "and downgrade to the lower version" add any value
here?  The word "downgrade" can have negative connotations in some other
contexts, so if it's not adding value I'd suggest avoiding it.

Section 8

   As reported in Figure 13, the values of the fields of the <extension>
   element (either new information or new messages) take the following
   values:
[...]
   o  the major standard version of the protocol that the extension
      refers to.

The XSL includes a full version (including minor), even though the
semantics basically only use the major version.  That said, why is the
'version' element minOccurs="0" -- what are the semantics when it is
absent?

Section 8.1

   The CLUE data model document ([I-D.ietf-clue-data-model-schema])
   envisions the possibility of adding this kind of "extra" information
   in the description of a video capture by keeping the compatibility
   with the CLUE data model schema.  [...]

nit: I don't think this is grammatical; maybe just "keeping compatibility".

Section 10

This claims to be a "call flow" example, but the described flows only
contain a single unidirectional media flow, which is not really consistent
with the normal usage of the word "call".  Buried in the body text there is
a disclaimer:
   [...]                      For the sake of simplicity, the following
   call flow focuses only on the dialogue between MP CP1 and MC CP2.
I would suggest making the presence of this simplification much clearer
from the start, perhaps "CLUE protocol messages exchanged in the following
call flow are detailed; only one direction of media is shown for
simplicity, as the other direction is precisely symmetric".

   CP2 acknowledges the second 'advertisement' message with an 'ack'
   message (Section 10.7).

   In a second moment, CP2 changes the requested media streams from CP1
   by sending to CP1 a 'configure' message replacing the previously
   selected video streams with the new composed media streams advertised
   by CP1 (Section 10.8).

This might be an appropriate place to indicate that media from the previous
configuration continue to flow during the reconfiguration process.

It might also be worth noting again somewhere in here (or a subsection)
that there are three (well, two, since we only show one direction of media)
distinct sequence number spaces per sender, and that the discontinuity
between Section 10.2 and 10.3's numbers is correct.

Section 11

Thanks for the well-thought-through security considerations; in combination
with the linked documents (particularly the framework), they cover all the
considerations (especially privacy considerations) that I had in mind.

Section 14.2

We appear to be citing both 5117 and 7667, whereas the latter obsoletes the
former.