Wednesday, July 9, 2008

Why network virtual machine

As a counter point to the Web Services Architecture and REST perspectives on modeling HTTP interfaces [1, 2, 3], the perspective of the Network Virtual Machine [1] emphasizes the atomicity of REST and works in the space of Interaction [1] to look at HTTP as a "web of peers" protocol.

Generally, peering protocols may push or pull messages. The IRC [1#Concepts] protocol demonstrates how servers can peer in a "web" by pulling channels from each other. However, unlike web HTTP, IRC channels are bidirectional command or instruction streams. HTTP can do this with a protocol upgrade [1#Upgrade], as in the case of TLS [1].

Current work on Interaction [1] includes work on multi node processes in terms of process algebras. Building up to processes in a more discrete way is a mechanical approach. Maybe it's not unlike looking at the subject of statistics in terms of counting.

The objective is to understand (comprehend) these processes with layers of theory founded in practical terms. The work on classes of process algebras for known behavior is important to representing relatively complex application protocols. While scalable internet services are best designed and implemented with much simpler interfaces.

Multi node processes over these network interfaces are more effective over simple protocols. Simpler interfaces are better Web APIs. And simpler interfaces perform better in the hosting system, for a better experience and lower cost.

No comments: