Re: Caching Servers Considered Harmful (was: Re: Finger URL)

James 'J' Allard (
Tue, 23 Aug 1994 02:40:04 +0200

| I suggest that the only reason that caching servers exist is to improve
| network resource usage, and by doing so, improve the responsiveness of
| retrievals. This is for the benefit of the caching server's network usage
| and against the best interests of the publisher.

i can't agree with this assertion. the benefits of caching are distributed
among users, sites, backbones, and the content server themselves.
the "publisher" of the ncsa home page would probably prefer that
large sites (read 10,000+ users) cached their page so that every user
loading up mosaic didn't have to hammer their server, their link,
their neighbors' link, etc. the publisher has the ability to limit the
duration this information can be cached, and should. i assert that
caching servers which don't respect the ttl are broken.

a http-based stock ticker service probably doesn't want to be
cached while the market is open, but when it's closed, why not?
the ncsa home page could probably stand to be cached for
a reasonably long period, as could many of the information
"leaves" in organizations' webs.

there seems to be a number of issues here that people are trying
to collapse into one:

1. copyrights
2. timeliness of documents/cache consistancy
3. network resource utilization

the copyright issue will only be solved when authentication
protocols are offered, and even then, cannot be contained.
information piracy happens independent of the distribution
channel for almost all information digital or print. this is an
issue that's been present in the internet forever w/ ftp mirror
sites. authentication will discourage the legality of caching
in some scenerios, it will be important that the notion of
anonymous or guest be permitted to allow caching to
continue on public, cachable information like homepages

different web sites, and documents within the sites are going
to have different timeliness qualities. it's important for
caching servers to respect these, and i think that work can
be done to help educate administrators, and make it easier
for them to control policy. ultimately, the end users will fix
any problem that this incurs. if my site admin caches for 5
days for all information, and i'm trying to use the stock
service, i will let my admin know that their caching server
is not acceptable and the matter will be corrected by
the admin, the servers' author, or the user (by bringing
their business to another provider)

caching is a good thing for resource utilization, it lessens
the burden on your provider, on the internet itself, and
on the resource you're trying to connect to. overall, it's
win-win when the content is fair-to-be-cached. sounds like
the big issue that's been raised here is the definition of
what is fair-to-be-cached.

proposals on cleaning up this definition and the
mechanism driving cachability of documents, or are
people satisfied in the conventions to date?
J. Allard
Program Manager of TCP/IP Technologies work: (206)882-8080
Microsoft Corporation home: (206)860-8862
"On the Internet, nobody knows you're running Windows NT"