The biggest shake-up in the underlying structure of the World Wide Web is on the way with proposals to change completely the nature of its underlying HTTP (HyperText Transfer Protocol), says our sister publication OnLine Reporter. The likely outcome? A move from HTTP’s simple ‘Request-Send’ model to one involving distributed objects. With work on specifying […]
The biggest shake-up in the underlying structure of the World Wide Web is on the way with proposals to change completely the nature of its underlying HTTP (HyperText Transfer Protocol), says our sister publication OnLine Reporter. The likely outcome? A move from HTTP’s simple ‘Request-Send’ model to one involving distributed objects. With work on specifying HTTP 1.1 nearing completion, the Internet Engineering Task Force HTTP workgroup is aiming to shut itself down by year-end. But it is likely that the workgroup will be replaced with another, examining proposals for HTTP-NG, a next generation object-based protocol. My personal bet is that the Corba-style of definition of distributed object interactions will dominate the thinking of advanced Internet protocols says Larry Masinter, joint chair of the existing HTTP workgroup. The main work on NG to date is based on a proposal from Simon Spero of Verifone Inc subsidiary EIT Inc. Conventional HTTP requires a browser to set up a new connection to the server for each request – for each image, or piece of text on a page – resulting in huge overheads. Next Generation attempts to reduce this by enabling multiple requests to a server to be passed over a more persistent single connection. Moreover, requests can optionally be handled asynchronously with the client able to fire out multiple requests without waiting for a response. Today’s browsers try to mimic this effect by opening multiple connections to a host, but this ploy tends to clog up router caches on backbone connections.
Multiple language support
While Spero’s work substantially cuts the network overhead and paves the way for better multimedia integration, the World Wide Web Consortium wants to go a lot further and bring object technology to the underpinnings of the Web. Moving HTTP towards an object framework would make it fit more comfortably into the corporate software world, and allow for multiple language support. A joint conference with the Object Management Group late last month was set up to explore ways in which the Web and objects could come together. The Web Consortium’s favored option is to implement a HTTP-NG proxy using Xerox Corp’s Palo Alto Research Center’s Inter-Language Unification architecture. Inter-Language Unification is a meta architecture that enables, so its proponents say, the easy bridging of multiple protocols with multiple language bindings. You want to link Corba objects to code expecting Sun Remote Procedure Calls? No problems, say the Inter-Language Unification guys. In a briefing paper, Web Consortium architect Dan Connolly enthuses: ILU makes all sorts of gateways and interoperability mechanisms nearly trivial. You want a gateway that translates between Sun RPC and CORBA IIOP? Just link both transports and message formats into your server, and you’re done. Eventually, I expect the same will be true of Distributed Computing Environment, HTTP 1.X, IBM Corp’s DSOM, Microsoft Corp’s DCOM and others. This will be critical when you want to run HTTP-NG from your palm-top, across a serial link to your personal computer, across the local network, through the firewall, and over the Internet, then back through a similar set of twisty passages to the server. No single message format and transport will be suitable for the whole route. Such a pragmatic approach is unlikely to endear itself to the Object Group, which wants to see IIOP everywhere. Ripping up today’s HTTP infrastructure, to replace it may seem like a pipe-dream – do they really want to replace 30m Web browsers? That is why, initially, the Web Consortium is working on proxy systems. End-users would be able to point their existing browsers at the proxy and garner at least some of the performance benefits of HTTP-NG as the backbone network of large servers gradually begins to talk the new language. The first browser to support the new protocol directly is likely to be Athena – the old work-horse developed by Dave Raggett, which is habitually used to test out new developments such as HyperText Mark-up Language 3. Work on the new version of Athena has yet to start.