Path: utzoo!attcan!uunet!seismo!rick
From: rick@seismo.CSS.GOV (Rick Adams)
Newsgroups: comp.sys.sequent
Subject: Re: BIND and resolver routines under Dynix 3.0.4
Summary: inadequate
Message-ID: <44375@beno.seismo.CSS.GOV>
Date: 14 Jul 88 21:52:12 GMT
References: <143@daitc.ARPA> <2961@cs.utexas.edu> <4080@oxtrap.UUCP>
Organization: Center for Seismic Studies, Arlington, VA
Lines: 31

You haven't lived until you've had to wait for a NS32032 processor
(even if you have 14 of them, some things don't work in parallel) do a
linear search on a 6,500 line /etc/hosts (thats how big the arpanet
host table is today).

Ignoring the issue of not being able to connect to sites not in the
host tables, the speed alone is enough reason to scrap the current
implementation (I'm sure that it runs fine on their 40 line host file
for their local ethernet). It's so bad that I've been begging them to
send us NFS so we can use the yellow pages!  (I loathe the yellow
pages. The yellow pages are a wonderful example of how a good idea can
be destroyed by a bad design and a worse implementation)

Our "solution" was the obvious one. For critical programs like
sendmail, we totally scrap what Sequent delivers and install a version
that uses bind.

For things like ftp or telnet, we have a small program that uses bind
to turn a hostname into an ip address. Then we ftp to the address.

The Sequent could be a wonderful networking machine if they would let
the engineers keep the networking software within 3-4 years of current
technology.

Sequent has a very nice hardware base, but I really wish I could
call the software something better than "adequate".

(We've had ZERO hardware downtime in 16 months if you need an
example of how the hardware works.)

--rick