On 4/15/2021 5:39 PM, John R Levine wrote:
On Thu, 15 Apr 2021, Christian Huitema wrote:
Adding test vectors would help, especially broken vectors.
+1. That would be a pretty good way for the IETF to help clean the
mess. That, and maybe a DNS site that would serve the test vectors.
In this case I think it's a reasonable idea but I echo jck's concern
that test vectors can turn into de-facto standards, particularly when
the tests and the text turn out not to exactly match.
A test server could easily have a fuzz option...
On the other hand, is it valid for a DNS compression pointer to point
forward in the message? Why or why not?
I had to implement a packet parser a few years ago, and I rather
naturally added a test that the pointers had to be backward. Not doing
that looked like inviting trouble. I tested that on a fair number of
root server traces and never found an issue. So I tend to think that's a
fairly safe test.
-- Christian Huitema
_______________________________________________
DNSOP mailing list
DNSOP@ietf.org
https://www.ietf.org/mailman/listinfo/dnsop