Stuart D. Gathman wrote:
Meng meant that good implementations would not hit the
recursion limit because they have loop detection, and would
stop after the 2nd level.
Yes, but the proposed limit 10 catches this too without much
overhead. Meng's loop detection:
0 - get policy of a.example (which includes b.example)
1 - mx, 2 - ptr, 3 - get policy of b.example
4 - mx, 5 - ptr, 6 - again a.example, exit with error.
Wayne's processing limits (steps 0..5 are identical),
6 - get a.example again (fast, it's still in the DNS cache)
7 - mx, 8 - ptr, 9 - get b.example again (now all cached)
10 - mx, 11 - exit with error, 10 is the limit in schlitt-01.
Bye, Frank