I did a cursory review of the Coman drafts.

These consolidate a rather useful body of knowledge, even if it is still in 
flux in many areas.
It may be more important to get them published in some state, with an explicit 
qualification of being somewhat work in progress, than trying to further polish 
them until nothing of substance is left.  My strong recommendation here would 
be: if in doubt, publish.

The most important aspect that probably needs to be expanded is the security 
considerations, with a view to resiliency.  Not having a management interface 
is a security liability.  Adding one is an even greater security liability.
Instead of trying to say what is needed here in my imperfect words, I’d like to 
refer to the discussion under number (5) in 
http://geer.tinho.net/geer.blackhat.6viii14.txt

In a related vein, many of the coman-specific management issues stem from the 
peculiar lifecycles the devices go through.
draft-garcia-core-security is the best document we have about this aspect, and 
the two coman documents probably should be developed in unison with this one.

It is not always easy to separate “constrained network” considerations from 
“constrained node network” considerations, and probstate might benefit from one 
more editing round to get the right focus in the right places. (Adding another 
unfocused term, LLN, and alternating back and forth, does not help.)  A related 
editorial problem is the usage “non-constrained” (which of course is physically 
impossible) — in RFC 7228 we missed coming up with good terminology for what we 
mean by less-constrained or affluent systems.
But these are more general editorial problems, and it is maybe too much to 
expect them being solved here.

I support Thomas’ comment on the use of the term “requirement”.
The definition of “Priority” in Section 3 of probstate, where a requirement 
suddenly becomes less important just because it happens to be impossible to 
fulfill, is a symptom of this.

Editorially, the requirements list would become less tedious if Requirement 
Type, Device Type, and Priority could be put into tabular form, maybe together 
with Req-ID.

2.002 does not mention the cost of copying and maybe should.
2.003 may be a bit optimistic about “compression” (as opposed to concise 
representation in the first place), but then this term isn’t really defined 
anyway.

A major problem with using classical network management thinking for 
constrained devices is that, for these, there may be little difference between 
network and application configuration, including firmware updating.

There are some actual “requirements” (6.*, 9.001 — which is “medium” 
priority???); it may be worth calling them out.
I don’t understand 10.004 vs. 6.* — while secure message transport is a good 
choice in many cases, it is ultimately there to support security objectives.

A great aspect of the use-case document is that it calls out the management 
responsibility for some of the use cases; this should always be done.  It would 
be good to know in which cases that responsibility is clearly separable from 
the application responsibilities (in the way network management has been 
traditionally).
“Medical Applications” is not a use case, it is a whole universe of use cases, 
and one or two should be included as their own use cases (e.g., separating a 
BYOD/personal interest case from a professional/healthcare use case).

I’m sorry for throwing out all these observations in front of the authors; 
again, if in doubt: publish, but maybe expand on some of these observations in 
a “future work” section.

Grüße, Carsten

_______________________________________________
OPSAWG mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/opsawg

Reply via email to