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

Rainer Klute (klute@tommy.informatik.uni-dortmund.de)
Tue, 23 Aug 1994 09:35:08 +0200


"Daniel W. Connolly" <connolly@hal.com> wrote:
>It's a simple matter of server configuration. As I recall, the CERN
>http daemon's default configuration is "correct" as per the protocol,
>i.e. it doesn't do any heuristic caching -- it only caches HTTP
>requests with an explicit Expires: header in the response. They do
>document the configuration options for heuristics like "cache FTP
>stuff for 2 days."

Not quite. The CERN server indeed implements a heuristic approach,
but a reasonable and configurable one. If a document comes along
without an Expires header but with a Last-Modified date, the cache
calculates an artifical expiry time. This is done by estimating the
document's age and adding 10%. For example, if a document has an
age of one month, it will expire after approx. three days. If it is
just one hour old, it will expire after six minutes. Thus, the more
frequently a document changes, the more frequently the cache will
check for a new version. You can also set an absolute upper limit
after which the server checks for a new version regardless of
expiry, you can modify the default of said 10% to some other
factor, and much more.

Dan already advised you to check the documentation. To put it
short: RTFM, folks! (That is, read the FINE manual!)

Best regards
Rainer Klute

Dipl.-Inform. IRB - immer richtig beraten
Rainer Klute EXUG - European X User Group
Universitdt Dortmund, IRB
D-44221 Dortmund Tel.: +49 231 755-4663

<http://www.informatik.uni-dortmund.de/~klute/>