> I am looking for a masters project on caching mechanisms
> for NCSA Mosaic to reduce network traffic and network latency.
One idea I would like to see explored is to make a start at handling
URNs. The basic idea goes like this:
a) Each object registered with the "network" is described by
a file containing meta-information for the object, e.g.
the author, expiry date, copyright holder, title, keywords
from a controlled vocabulary etc. The registration process
generates a globally unique name (a URN). Mosaic's Fill-out
forms seem like a natural interface for registration.
b) Web servers cooperate to provide a distributed name look up
scheme similar in spirit to DNS, *BUT* which allows for
distributed copies of objects as well as naming info. The
caching hierarchy is modelled after DNS, following the
IP network/subnetwork divisions.
This approach saves time by combining the process of resolving a URN
to a physical address and retrieving the object. Given that you are
after a masters degree, I think you should aim to keep things real
simple. The idea introduces server to server communication and could
be built by adapting existing server software. Don't worry too much
about the syntax for URNs; I would recommend something like:
gun: name [.name]* e.g. gun:mybook.me.mydepartment.myuniversity.edu
I have used gun for "globally unique name" :) to avoid confusion with
the ongoing IETF efforts to come up with a consensus for URNs. This
choice would fit the globally unique names into the URL name space.
Let me know if you want more detailed info on how this all would work.
I have lots more ideas, e.g. for rule based binding of URNs to objects
based on properties of the client, aliases, non-hierarchical naming
schemes using centroids, and pre-emptive caching.
-- 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