On Mar 30, 2005, at 12:13 PM, Dave Crocker wrote:
it is probably worth distinguishing between a) behaviors that will
problems with typical, current implementations, versus b) behaviors
cause problems inherently due to the design of the DNS. Caching
well might be alleviated by different implementation schemes and/or
different key/name administration.
Having to change the implementation is, of course, significant, but it
quite different from breaking core service design limits.
I suspect the answer is C) none of the above. As for (A), I don't
think too many implementations fall into this category, and the ones
that do should probably be fixed for other reasons. As for (B), the
trade off is increased cache size vs. increased query load; no amount
of fiddling with the DNS specs is gonna get you around that.
None of this means we shouldn't move forward. We should probably note
DNS load issues and cache sizes as an operational consideration and
keep moving forward.