I read it, and I think it's OK to publish, but I found it very
frustrating. It describes a zillion options but it doesn't tell me what
to do. I realize that there's a fundamentally intractable problem when
the info you want to bootstrap your secure DNS channel has to be looked up
in the DNS, and also realize that since this is so new, we have no
operational experience so the most we could do would be a BWSP.*
You might consider adding something about the info leakage tradeoffs
of various bootstrap schemes. For example, if the client knows the
name of the cache it wants to use, and queries in the clear for SRV,
A, and TLSA, those will leak but as soon as it connects to port 853 of
the cache, anyone sniffing traffic will know that's the cache it's
using anyway, and anyone with passive DNS could recover the name. Or if
you distribute the name and DNSSEC chain out of band, that hardly leaks
anything but has all the problems of manual key distribution.
Once we have experience, a real BCP describing what works in practice
would be useful, too.
Micronit in sec 6.5 third para, "at least on piece" -> "at least one piece"
R's,
John
* - Best Wildly Speculative Practice
_______________________________________________
dns-privacy mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dns-privacy
