on TTL expiry BIND sends 'ANY' query, gets back 'NOANSWER'

[Phil Mayers]

On 14/04/15 00:44, Mark Andrews wrote:

> No.  Named caches NXDOMAIN and NOERROR NODATA to ANY queries
> indendently of qtype (with the exception of DS/NXDOMAIN).

Shrug. As I've said a couple of times, I'm not experiencing this 
problem, so it makes no difference to me. I'm really just wondering aloud.

> I'm getting tempted to remove the work around code for non response
> to EDNS queries.  I'm also tempted to remove the ability to say
> that EDNS is not supported in named.conf.  Named will still fallback
> to plain DNS on FORMERR and NOTIMP.  Yes, this will break lookups
> to certain zones.

If you think that will help in the long term, then it might be a 
reasonable thing to do.

I'm deeply sympathetic to the frustrations of an implementer faced with 
other lazy implementers, and the moral hazard in working around other 
peoples bugs.

That said, I can empathise with the likely resultant grumpiness - people 
who had working resolvers now don't, and for a "boring technical reason" 
that they really don't care about.

It's a sorry state of affairs that we're still fighting basic protocol 
compliance issues in 2015. I was supposed to have a flying car and 
hoverboard by now...

DNS is an old protocol. I wonder how long we can expect it to work 
without a compatibility break. 50 years? 200? 1000?

> Using EDNS extensions will be the next battle field.  There are

I think describing this as a "battle field" might be framing the problem 
in a rather unfortunate way, but that's just my opinion.

Cheers,
Phi