disregard previous whinings...

Hi,

I think I only bothered select people from Merit and ISI about this,
but I forgot who they all were and rr-impl out to cover the list and
more.  I had previously noted (OK whined about) the
agen/peval/radbserver/ripe combo being incredibly slow and the problem
was right in front of my nose and I didn't see it.

I had made some changes to speed up enparse that avoid massive string
concatenations on the mega- aut-num for AS690.  This made dbupdate
tolerable from an insertion time standpoint.  whoisd and radbserver
still ran real slow.

Somehow I either never changed or manage to undo the change to enread
to accomplish the same thing, even though the change to enread to
avoid passing the object by value was still there.

Oh well.

Bottom line is:

  time whois -h figaro ' -s anstest as690' > /dev/null
  0.6u 0.4s 1:16.64 1.4% 0+0k 0+2io 0pf+0w

This was taking 36 minutes on an idle system and I timed this on a
system with CPU usage already pegged at 100% by radbsaerver at the
1+1/4 minute above.

My appologies to anyone that spent any time looking into this.  Now
back to why radbserver is so slow after enread comes back...

Curtis


ps - full patches to follow when I get things in order.



-------- Logged at  Tue Jun 20 00:55:48 MET DST 1995  ---------