On Feb 15, 2010, at 1:01 PM, Seth Mattinen wrote: > On 2/15/10 9:21 AM, Tony Finch wrote: >> On Mon, 15 Feb 2010, Mark Scholten wrote: >>> >>> I've seen problems that are only there because of DNSSEC, so if there is a >>> problem starting with trying to disable DNSSEC could be a good idea. As long >>> as not all rootzones are signed I don't see a good reason to use DNSSEC at >>> the moment. >> >> You realise that two of them are signed now and the rest will be signed by >> 1st July? >> > > > Which means now is a good time to find and fix brokenness, not hope that > DNSSEC will go away.
Right. Apart from implementations that just can't handle funky RR types in the response -- firewalls, perhaps? see RFC 2979, especially the transparency rule -- a lot of the trouble is caused by the reply size. The code should either use EDNS0 or fall back to TCP -- and lots of folks have broken firewall configs that don't allow TCP 53, even though it's been in the spec since 1984 or thereabouts. --Steve Bellovin, http://www.cs.columbia.edu/~smb