Re: Caching Servers Considered Harmful

Dave Raggett (dsr@hplb.hpl.hp.com)
Tue, 23 Aug 1994 17:44:15 +0200

Rob Raisch writes:

> The publisher holds complete responsibility over their product, in
> content, presentation, timeliness and distribution. By running a caching
> server on my content, you are robbing me of any control I might have over
> the timeliness and distribution.

Timeliness can be managed via including information on when the information
expires and pointers to routes for obtaining asynchronous notification of
changes. The latter is something that can be handled separately from HTTP,
e.g. you could use email or a reliable multicast protocol.

The scaling issue will mean that publishers do indeed lose control of
where their information is being held in the net. The real problem is
ensuring that only people who have paid or are otherwise entitled can
actually read the information, no matter where it is held. This is a
challenge to the security folks and I can see a number of ways to handle this.

In the longer run we need to sort out a model for what kinds of meta info
are needed to describe objects, e.g.

o Can the object be held in a distributed cache with automatic
replication (airline booking can't be done this way).

o Info describing how URN requests bind to objects, e.g. rules
for binding depending on geopolitical cues, such as which
country or organization the client belongs to. Binding may
also depend on time, such as the latest issue of a magazine.

o Expected lifetime of the resource.

o How to hook/unhook into automated notification of changes
and whether this is a requirement for holding copies.

o How to get authorization to view the object. This includes
consideration of payment schemes.

o Publishing info such as author, date and organization

o Info for supporting searches for information, e.g. lists
of keywords from a controlled vocabulary.

Given this kind of info, we can then get URNs and distributed caching
to work, something that will become increasingly important for
dependability as the Web continues to grow exponentially.

--
Best wishes,

Dave Raggett

----------------------------------------------------------------------------- Hewlett Packard Laboratories email: dsr@hplb.hpl.hp.com Filton Road tel: +44 272 228046 Stoke Gifford fax: +44 272 228003 Bristol BS12 6QZ United Kingdom