APIs - was Re: Visibility

Mark Baker wrote:

> On Wed, Feb 19, 2003 at 08:40:14PM -0700, Champion, Mike wrote:
>
> >>So where are the standard Web service APIs?
> >
> >Huh? APIs???
>
>
> Yah, APIs.  HTTP provides a standardized one.


Oh really? *ONE*?!?! Where, pray? In what language? C? C++?
Java? C-sharp? Python? Fortran? Visual Basic? [continued on page 94]
And "standardized" by whom?

I hope you're not referring to libwww, because http://www.w3.org/Library/
explicitly states that "The purpose of libwww is to serve as
a testbed for protocol experiments." Nowhere is it claimed
that it's any kind of standard - even for C.

> Why haven't I seen people talk about API standardization yet?  Is it
> considered a good thing that if Altavista wants to Web-service-enable
> their search engine, that they should use a different API than Google?
> I would hope not.

API standardization and protocol standardization serve two quite
different sets of objectives. Protocol standardization enables
interoperability and (sometimes) observability. API standardization
enables code re-use and platform consistency. Inevitably (and with good
reason) there are likely to be many APIs: in different languages, optimized
for different environments, with different dependencies. The functional
and performance characteristics required in a cell-phone app are
not the same as for a mainframe, and the resources available - including
supporting APIs - are likely to be quite different in the two cases.

Over the years I've come to the conclusion that the skill sets required
for protocol and API design are sufficiently different that it's a mistake
to have the same people do both. And in practice they rarely do - most standards
bodies emphasize one or the other. The technologies, economics, and yes, politics
involved are quite different.

That trout is a red herring. Out of the pond with it!

Geoff

Received on Thursday, 20 February 2003 00:27:01 UTC