On 03/12/2025 17.21, Petr Špaček wrote:
Indeed a good question! Originally I was thinking this would apply
only to SERVFAIL answers and TC=1 generated by RRL is not the case.
Is there a use-case for 'soft' vs 'hard' limit distinction?
Hard limit exceeded = SERVFAIL, go away.
Soft limit exceeded = TC=1, you might want to try again?
I don't know. WDYT?
I haven't thought about it in depth, but I assume the question is for
everyone, so hopefully we'd get more reactions.
It might make some sense to express how much ephemeral this limit is,
though there will surely be some gray areas, e.g. limits not getting
exceeded after more things get into cache (a hard to predict case,
moreover).
I can e.g. imagine that forwarding code might like to utilize
information that it's sending too many queries (to this IP?) and perhaps
try to prioritize other IPs in the configured set. In case of
forwarding, latency isn't such a great source of information for
load-balancing (maybe latency of cached replies, but...)
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]
