RE: Genericity, strong typing, SOAP-RPC, GRASP, and defining the Web

> Types don't merely have complexity costs, they have complexity 
> benefits as well. A type is a constraint: it sets bounds on the range 
> of acceptable values of any instance of that type, and as such 
> _reduces_ complexity.

'acceptable' seems slanted toward the receiver of the value.
Given 10 receivers - each with different constraints - the complexity at
each receiver might be low, but the complexity at the sender is 10x.
How many senders does a receiver want to play with? It depends on the
attitude toward incoming messages :
If a receiver says "I don't care, as long as everybody sends stuff in my
format" then senders will see a myriad of choices and have to slog through
building many adapters, and the network will be quiet for a long time.
If a receiver says "I party with anybody" then the network will be busy

You can see that on the Web today - each FORM POST is a unique format, and
almost nobody cross-posts to other sites (to 'compose' new services, etc.).
However, each GET is a consistent format and almost everybody (can)
cross-link to other sites.

If XML-Protocol is a new application protocol, then it should have a new
port and uri schema: xmp:// or whatever.
If it is not a new protocol, but people want to use it with HTTP, then the
xml data description language portion of XML-Protocol should be a registered
mime type for use in Content-Type and examples should be given for how to
use it with GET, PUT, POST, etc.

Received on Tuesday, 26 March 2002 19:55:02 UTC