Re: Caching violates copyright?

On Thu, 14 Aug 1997, WWW server manager wrote:

> > > Every website that uses an out-of-the-box configuration, like Slackware
> > > has this turned off by default..
> > > Personally i feel that this option SHOULD be ON by default !

Personally, I feel "Expires:" headers should be on by default...

> required). The only time it would be sensible to allow caching of such
> documents is if you (as origin server admin) know that although the documents
> could differ, the variants are sufficiently similar that it wouldn't really
> make much difference which version someone received. (In which case, it's
> not obvious that there would be any point in supporting multiple variants.)

Or if the negotiation was simple enough (could be determined from response
headers) that Squid could actually return the "right" document from its
cache, even if the URLs matched? Maybe something a few versions down the
track...

> cached copy of *that* variant. Simply applying variant selection criteria to
> any variants that happened to be in the cache would not be sufficient, since
> the cache could not know whether the origin server could offer a vastly
> better match to the user's requirements than any of the variants that
> happened by chance to be in the cache.

True - but aren't we playing that guessing game already with caching? I
mean, a Web server may indeed send "Last-Modified" and "Expires" headers,
but it may STILL (for whatever reason, good or bad) send different output
for each request - yet the proxy will still cache the result for a while.

> For example, consider a request specifying that the user would prefer the
> document in Icelandic, but would accept English if necessary, and the only
> cached copy was in Russian. Returning the Russian version or rejecting the
> request as impossible to satisfy would both be unreasonable.

The cache should see that the only copy is in Russian and therefore treat
this as a cache miss and proceed as it would normally do in such a
circumstance.

Obviously if the server doesn't use content headers properly then this
could make things tricky, but if we keep allowing Web server authors and
administrators to make these mistakes, will we ever get technology such as
full content negotiation into widespread use?

> If the cache also had the English version, returning that would be more
> reasonable, but still giving the user poor service when the origin
> server could have provided precisely what had been requested.

But if the remote server could not be contacted, then returning the
English version would be better than an error message. ;-)

Cheers..

dave
Received on Sat Aug 16 1997 - 00:53:23 MDT