I've analyzed my DNS cache logs in more detail:
* The number of different MX targets is 97% of the number of
different MX records. Weighting by query doesn't change the
figures.
* The number of different address sets for those MX targets is 94% of
the number of different MX records. Again, weighting by query
doesn't change the figures.
* 40% of all MX targets are outside the MX server's bailiwick. The
cache can't use the A even if the server provides it.
In other words: There's very little overlap between mail exchangers. The
number of queries wasted by missing addresses is much larger than the
number of queries that could possibly be saved by overlap.
Elz is wrong when he blames the lack of overlap on the choice of MX
names. In fact, almost all of the addresses are different.
Elz concludes, from his incorrect efficiency claim, that my advice to
avoid gluelessness is ``bogus.'' Apparently he doesn't understand that
gluelessness causes _reliability_ problems. I explained this in detail
in the A6-unreliability thread.
---Dan
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page: http://playground.sun.com/ipng
FTP archive: ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
