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