In message <56E768E6(_dot_)5090905(_at_)dougbarton(_dot_)us>, Doug Barton
writes:
On 03/14/2016 04:18 PM, Paul Wouters wrote:
Yes, you are about 1.5 years late. And your arguments are (un)fortunately
not new arguments. Since the archive on this draft is rather huge, I can
understand that you missed part of this discussion. So for completeness
sake, I will answer your questions again.
Thank you for your patience in explaining your reasoning, and again, I'm
sorry for coming late to the party. And thanks as well for confirming
that my memory is correct ... at one time I did hear that this topic was
going in the direction of signatures rather than certs. Unfortunate that
I didn't follow it closer.
Regarding what you said and what your goals are, I think that we are
pretty far apart. I will send a detailed response to your message on the
DANE list soon. In all likelihood I will also create a new I-D with my
ideas specified in more detail. Perhaps what is needed is more than one
experiment. :)
In regards to the current last call, while your explanations do help to
alleviate a few of my concerns, in large part I am still not
enthusiastic about this version of the draft proceeding.
In particular I think the concern about these RRs being used for DDOS
amplification remains. There is no mechanism in place currently in any
name server software that I am aware of to limit responses to queries in
the manner you describe (only send answers if the query comes over TCP
or with DNS-Cookies). Further, I don't see that happening any time soon.
You just limit response sizes in general. BIND 9.11 has
"nocookie-udp-size <integer>;" which sets a EDNS response size limit
for queries w/o a valid server cookie. If the response doesn't fit
you do the normal fallback to TCP. With EDNS both sides can set
limits on what they are willing to send/receive.
Amplification controls should be independent of qname and qtype.
Close behind that concern, the larger IETF community (or at least some
very vocal segments of it) have serious concerns about this type of
opportunistic encryption happening at all, or in my case, without user
input. They (and to some extent I) remain unconvinced that your
assertion that this type of opportunistic encryption is always better
than the current state. Personally, I need to think more about that, but
at least in the early stages of an experiment in tying PGP keys to DNS
RRs, I'm definitely opposed.
FWIW,
Doug
--
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742 INTERNET: marka(_at_)isc(_dot_)org