DNS and Mosaic

andym@inrird.com (Andy Micone)
Message-id: <m0oiu1l-00000uC@inrird.com>
From: andym@inrird.com (Andy Micone)
Subject: DNS and Mosaic
To: www-talk@nxoc01.cern.ch
Date: Fri, 1 Oct 1993 16:36:44 -0700 (PDT)
X-Mailer: ELM [version 2.4 PL21]
Content-Type: text
Content-Length: 1743      
Marc Wrote:
"DNS hostname lookups often take a very long time in relation to the
actual time needed for the
connection and data transfer. should we therefore maintain a cache of
hostnames and IP addresses on the client side to try to paper this
(non-Mosaic) problem over?"

Well my first reaction is don't do it, because any Administrator who
doesn't run DNS so it caches  addresses is just asking for performance
problems. We, for example, connect to the internet through a 14.4K
modem, it's pretty slow. Hostname lookups, however, aren't one of our
problems since one machine on the local LAN caches address lookups.
It's the Administrator's job to resolve log-jams like this, at least
that's my philosophy.

It is, however, off-putting for a user to have to wait to for anything
to come up on a computer. In the real world you can just turn to the
footnote in the back of the book, or stare down at the bottom of the
page. A computer lookup of a link should take no less time. The
acceptance of a technology is often based on how easy it is to access,
and a user does not care, nor is knowledgable about how things happen,
they just should happen as quickly as possible.  The web is an
important technology, and I think people working with it should make
every effort to speed its acceptance. Further, some users may be using
a slow link to the Internet through SLIP on non-UNIX machines, so
local DNS isn't an option for them. 

I conclude, therefore, for non-Unix versions, definitely cache,
because its more than likely there will be no caching options locally.
For Unix versions it's a toss-up, my guess is that fewer people would
gain by its inclusion than would benefit by having DNS setup correctly
for their local cluster.

-- Andy