Hi Stuart,
Somehow our discussion has gone awry, and I'm not quite sure why,
because I am not sure that we fundamentally disagree with each other.
At least, I think that we both see some of the same potential
problems, even if we disagree about what steps would be appropriate
to resolve them.
[Please note: In this note, I will use the term "LLMNR" to refer to
the DNSEXT WG draft named draft-ietf-dnsext-mdns-XX.txt, and I will
use the term "mDNS" to refer to the de facto standard Apple mDNS
protocol.]
Other than a few minor issues that are being dealt with in a -43
update, I don't think that anyone has raised a blocking technical
issue with the LLMNR specification during this IETF LC. If you (or
anyone else) has intended to raise a blocking technical issue, either
with LLMNR itself or with its ability to coexist with mDNS, please
make that clearer to me.
I do, of course, see overlapping purpose/applicability for the mDNS
and LLMNR mechanisms, and that is a significant source of concern.
In general, I don't personally think that the Internet would be
best-served by having two different widely-deployed mechanisms for
link local name resolution. I can infer that the DNSEXT WG agrees
(or agreed at one time), since they decided that a selection process
was necessary rather than just publishing both (or all three)
mechanisms. So, I'm not quite sure what to do given that mDNS _is_
widely deployed, and that we are considering the standardization of
LLMNR in an Internet where mDNS must considered as part of the
existing landscape.
I am somewhat concerned, for example, that someone could implement
LLMNR thinking that it is the same thing as mDNS and only later find
that the two do not interoperate. Or that some vendors will ship
LLMNR while others ship mDNS, so that products from different vendors
will not interoperate. Do others have any thoughts on whether the
publication of LLMNR is likely to cause confusion along these lines?
On the other hand, the DNSEXT WG has worked for several years to
produce the LLMNR specification, and I don't see anything
fundamentally wrong with the mechanism that we have produced (people
should respond to the IETF LC if they see blocking technical issues).
The authors of that specification gave change control to the IETF
community, and they have gone through 40+ document iterations,
working towards a document that would achieve DNSEXT consensus. That
process was not followed for mDNS (because it was not the chosen
solution), and we currently only have one document (LLMNR) that has
reached IETF WG consensus and has been submitted for standards
publication.
It is possible, in an IETF LC, that we could learn that we do not
have IETF consensus to publish something that was produced by an IETF
WG. That would be an exceptional and unpleasant situation, but I am
open to that possibility. In this particular case, though, I see
that a few concerns have been raised that were already raised during
the WG process, and I do not see an IETF consensus against publishing
the LLMNR document that would override the DNSEXT WG consensus to
publish it. If folks disagree with this determination, they should
speak up, because I am currently of the opinion that we do have
DNSEXT WG and IETF consensus to publish LLMNR on the IETF standards
track.
Some folks have already implemented LLMNR, and at least one company
has shipped it in a commercial product.
So, I think that, no matter what happens, we are stuck in a world
where there will be at least two link-local name resolution
mechanisms. Having come to that conclusion, I don't believe that
there is any perfect answer here. I think that our focus needs to be
on minimizing the potential damage or confusion that will be caused
by having two mechanisms.
Stuart, have you considered publishing mDNS on the individual
submission track (as an Informational RFC)? I think that the
existence of that document might help to avoid confusion about
whether the LLMNR document describes mDNS.
The LLMNR document already includes an informative reference to
draft-cheshire-dnsext-multicastdns-05.txt, but it does so rather
obscurely (citing a need to use a TTL of 255 for "compatibility with
Apple Bonjour"). It might be better to include a pointer to this
document earlier in the LLMNR spec and indicate that Apple Bonjour
includes another mechanisms for link-local name resolution. Also, I
am surprised by the reference to compatibility with Apple Bonjour
(which I didn't notice in my AD review), because I was not under the
impression that these protocols are compatible (as in interoperable).
LLMNR authors, could you perhaps elaborate on what was meant by this
statement?
Would the LLMNR authors and or the DNSEXT WG object to the idea of
including an informative reference to mDNS (presumably as "Apple
Bonjour") earlier in the document, stating that it is another
link-local name resolution protocol and making it clear to what
extent LLMNR is (or is not) compatible with mDNS?
Thoughts?
Margaret
_______________________________________________
Ietf@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ietf