FW: BIND limits and performance questions
kcd at daimlerchrysler.com
Fri Mar 23 20:38:22 UTC 2001
TCP retransmissions are wasteful even when the software works perfectly.
I can certainly understand the impulse to not pamper -- maybe even
deliberately break -- non-conforming software. But wasting the Internet's
finite resources with unnecessary TCP retransmissions seems to me to be
cutting off one's nose to spite one's face. The cure could be worse than
the disease. (Hmmm, that's almost a mixed metaphor. I'll kludge it by
saying that the disease in question is nasal gangrene :-)
Maybe when EDNS0 is widely deployed, such a "non-conformance purge" (aka
"Great Leap Forward") might become practical...
Derek J. Balling wrote:
> > In any event, suffice it to say that many programs simply flake
> >out in the face of truncation, and you really need to make certain
> >you do everything you possibly can to ensure that you never cause
> >truncation in any form.
> ... and, of course, the more we pamper broken implementations, the
> longer they stick around.
> If, all of a sudden, some (or many) largish sites (let's say,
> Hotmail, Yahoo, MSN, Juno, AOL, etc.) was to start advertising 25 MX
> records, pissing all over the 512-byte issue, and say "Hey, its a big
> record, if you're not dealing with it properly that's your own
> business".... how long would sites continue to run old/broken code
> when they couldn't talk to -- or possibly even SEE -- large portions
> of the net?
> If more people were willing to say "you don't obey the spec, you
> don't get to talk to me", those breaking the specs wouldn't break
> them for all that long.
More information about the bind-users