[dns-operations] resimprove and Re: DNS Flush Protocol

[Paul Vixie]

Edward Lewis wrote:
> I just don't see using the parent TTL's as a way to go.

for scale, i've proposed a convention that requires no new transactions
and no new relationships. the parent TTL is communicated by the
authority server to the recursive server as part of delegation, but is
almost never used, because the descendant's apex NS TTL has a higher
credibility.

i see the fact that legitimate NS updates will cause cache purges as a
very minor issue. RDNS restarts also cause cache purges. LRU is the most
frequent cause of cache purge. if you were looking to optimize for the
average case, then the average case is that most data is not held until
its TTL expires.

if you have an alternative in mind that uses some other existing
transaction (like the delegation's TTL) then it could scale to the size
of the current and future internet, and in that case i'd like to hear
your thoughts.

-- 
Paul Vixie