Web Services: Its So Crazy, It Just Might Not Work

Clay Shirky clay@shirky.com
Wed, 1 Aug 2001 10:31:39 -0400 (EDT)


> I would argue that while, yes, there are significant coordination
> costs that result from an API approach, these are insignificant by
> comparison to the effort, time, and expense necessary to introduce
> new semantics into a protocol such as HTTP, 

To be fair, I don't think Mark wants to introduce new semantics into
http. I think he wants all the semantics that can't be encoded into
http methods to be put in the places http can carry other data.

This, I think, makes the difference between his pov and the XML-RPC
camps pov very slender, because once you decide to use POST as a
general purpose method, you've practically got XML-RPC.

> *even assuming* one's technical competence and conscience can
>  stomach such an attrocity.

Hey now. We've managed to have a pretty civil debate so far. This
business of conflating "I wouldn't do it that way" with "That way is
atrocious" is part of what keeps the real issues from surfacing.

> And as for working toward new, separate application protocols ---
> forget it.  

Channeling Mark again, I don't think he wants a new wireline
protocol. I think he wants people to use http straight, no chaser.

> The era of the wireline app protocol is over. 

This is interesting. Why is that era over, and when did it end?

-clay