A method of and apparatus for accelerating data retrieval on a widely
dispersed content distribution network such as the World Wide Web. The
system associates a concept URI with a resource which can then serve as a
reference for a group of resources. When the system receives a request
from a client calling for a resource in this group, it prepares a delta
encoded responses based on the difference between the reference document
and the response it has obtained from an origin server (the identity
response). The system then sends the client the delta encoded response, an
identification of the reference document used to prepare the delta encoded
response, and an address where the reference document may be obtained if a
client does not already have a copy. A client either decodes the
information from the system by reconstituting the response using the delta
encoded response and the reference if it already has a local copy, or, if
it does not already have a local copy, using the address to retrieve a
local copy and then carrying out the decoding. The selection of a
reference is based on heuristics. The system also has the capability of
changing which resource it will use as a reference based on the
performance of the reference, i.e., whether the reference has continued to
produce acceptably small deltas. The reference may be a version of an
actual resource, or created by the system to serve the group of resources
as a reference.
