[dns-operations] Truncation query
ray at isc.org
Mon Feb 6 15:22:10 UTC 2017
On 06/02/2017 15:13, Paul Vixie wrote:
> in answer to your first question, yes, ancount=0 would have been
> correct. you'd be within your rights to treat this as a format error,
> but it's probably more constructive to at least also re-try with tcp.
In this particular case my app is doing conformance testing, hence why
this is a question of "legality" rather than "what to do next".
Way back when I did SSAC035 I saw lots of very bad behaviour from
middleboxes, with most of them simply truncating the packet at the 512
byte boundary (with our without setting the TC bit).
I even saw a box that set TC=0 even if the upstream DNS server had
replied with TC=1 !
More information about the dns-operations