Well, the reality will be that DKIM is not alone in the new
dawn of dumping/exposing server policies in TXT records, i.e,
SPF, SENDERID and now DKIM and I'm sure others to come.
They don't count:
* Their records are much smaller (no 1024 bit keys)
*They are not standards track.
What Andy is pointing out here is that these are issues that have to be
addressed in the spec because they have operational impact elsewhere.
There may also be DoS attacks through cache exhaustion.
It is much better to get these issues sorted out now than wait till last
call or the IESG queue.