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

Daniel W. Connolly (connolly@hal.com)
Mon, 22 Aug 1994 12:28:44 -0500


In message <Pine.3.85.9408221227.A462-0100000@enews>, "Rob Raisch, The Internet
Company" writes:
>
>On Mon, 22 Aug 1994, Daniel W. Connolly wrote:
>
>> In message <Pine.3.85.9408221141.A462-0100000@enews>, "Rob Raisch, The Inter
>net
>> Company" writes:
>>
>> >You can provide no guarantee that the versions that you present to your
>> >users are accurate or timely.
>>
>> This is extremely misleading, if not just plain incorrect.
>
>Dan, how then do you address John's comments regarding the Dilbert comics?
>
>On Mon, 22 Aug 1994 18:53:24 +0200, John Labovitz <johnl@ora.com> wrote:
>
>>(From the paper, the HENSA server will expire GNN non-GIF files
>>after two days, and GIF files after two weeks. Here's a real-life
>>ramification of caching: for those using the HENSA server, our daily
>>Dilbert comic strip is available only once every two weeks.)

Simple: the HENSA server is broken.

You wrote:

>> >You can provide no guarantee that the versions that you present to your
>> >users are accurate or timely.

I took the "You" in that sentence to mean you, the caching server
administrator. It is simply not true that the HENSA server
administrator _can not_ provide a reasonable guarantee that his
readers get timely information. This is why I spoke up: so that
www-talk readers don't come away thinking what you said is the only
way it can be.

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."

I think it's even possible to say "cache GIFs for two weeks, except
for the dilbert comics from ora.com," if you're so inclined.

The point is: please don't generalize certain broken installations and
implementations into a condemnation of the WWW architecture as a
whole. I agree that WWW has a long way to go before it is consumer
technology, but perception has a lot to do with it's deployment. Let's
not give the wrong impression at this critical stage.

Dan