Re: fix mbuf exhaustion when flooding under ppp-on-demand

["William Allen Simpson" <wsimpson@greendragon.com>]

> From: "Angelos D. Keromytis" <angelos@dsl.cis.upenn.edu>
> I'd appreciate it if you could test this under normal ppp on demand,
> and then with a ping flood.

Nice try.  Unfortunately, wrong fix.  This stuff has been discussed long
ago on e2e-interest and tcpi.  I'm amazed that kind of thing is still in
any unix pppd.

1) For most link speeds and setup times, no packet at all should be
   kept!  Completely throws off the RTT calculation, especially with
   timestamp/paws.

2) When link setup takes less than a few milliseconds (ISDN), it is OK
   to keep a packet, but it should be a _random_ packet, neither
   head-drop or tail-drop.  See the RED manifesto.

3) In most cases, tail-drop is preferred to head-drop (which you did).

WSimpson@UMich.edu
    Key fingerprint =  17 40 5E 67 15 6F 31 26  DD 0D B9 9B 6A 15 2C 32