keepalive/lookahead

I had an interesting conversation with a friend who worked at Gain on
a multimedia/hypermedia authoring system they built several years ago.
(Gain is part of Sybase now).  He doesn't know much about the web, and
so is in a good position to make fresh observations.

I was telling him about the keepalive stuff that is happening now, and
he mentioned a use for it that is similar to something in the Gain
system but that I have not yet seen discussed in the context of http.
(It probably has been -- I just don't know any references, sorry).

Anyway, the Gain system had some notion of property lists on objects,
one of which was a property that allowed other objects to be
pre-fetched whenever a given object was fetched.

The http analog for that would be headers and/or HTML tags of some
kind that indicate other documents that should be fetched whenever a
given document is fetched because of the high probability that these
other documents will be viewed whenever the first one is viewed.  This
doesn't strictly require keepalive, but it interacts well with it, as
prefetch would be much cheaper to implement using it.  In fact, for
sufficiently small documents it would be statistically cheaper to
prefetch, in one TCP connection, some documents that go unread than to
set up and tear down TCP connections for every document that is read.

Sorry if this is too speculative for this group -- just thought it was
an interesting idea.

--Shel

Received on Monday, 28 August 1995 17:07:39 UTC