WWW Non-server computation

Steven Pemberton, Steven.Pemberton@cwi.nl

Currently on the World-wide Web, a client requests data from a server, the server delivers the data to the client, and the client does the work of presenting it to the user. This is the most equitable distribution of work possible.

However, when it comes to computation, the only model is that the client requests a calculation, and the host does all the work, replying with the results of the computation.

For a task like searching in large databases, there is no alternative to this, but for other computational tasks, this can be an unequitable use of resources; it puts an unreasonable load on the server, is not scalable, and consequently discourages the provision of such services.

What is needed is a model of computation that allows the client to do the lion's share of the work. Clearly, there needs to be a machine-independent notation; it should not be bulky, and it should offer guarantees of security (even PostScript apparently offers a potential security problem if it is interactively displayed).

As a computational model, a model based on previous research by the author with Lambert Meertens, the ``Views model'' of objects linked by invariants can be used. Site-bridging is modelled by an equality constraint, corresponding to the notion of proxy (or surrogate) object in object-oriented distributed operating systems. In the WWW context, URLs are effectively proxies. In this model data is no longer essentially static, which is a radical departure from the present WWW, and gives possibilities for dynamic and interactive documents.