[OAI-implementers] protocol comments, OAI 2.0

Walter Underwood wunder@inktomi.com
Wed, 30 Jan 2002 21:07:54 -0800

--On Wednesday, January 30, 2002 11:37 PM -0500 "Michael L. Nelson" <mln@ils.unc.edu> wrote:
> I agree with your point about a one-off protocol.  However, as Simeon
> mentioned, OAI 2.0 will not have http dependencies.

SOAP does not have HTTP dependencies. HTTP is only one transport
for SOAP. E-mail is another.

> Perhaps the best description is the primary goal of keeping OAI as simple
> as possible for data providers, even at the expense of service provider.
> The notion is that if being a DP is as *absolutely simple* as possible,
> more DPs will exist.  A critical mass of DPs will cause SPs to emerge.

Custom protocols are never simple. They are dead ends. There won't be
an O'Reilly book for OAI. There will be for SOAP.

Why build something and wait for the users to come? Take it to them,
in their language. Now. Even AppleScript can make SOAP calls. Really.

> So, I guess the next question is "who will write the first OAI <-> SOAP
> gateway?"

My question is, "why should it be neccesary, when it could have
been a SOAP protocol in the first place?" Why design something
that is destined to be an epicycle in the cosmology?

I'm guessing that 2.0 is "too far along" to change. That is not
a good reason to do the wrong thing. The model is already there,
recast it in WSDL.

Walter R. Underwood
Senior Staff Engineer
Inktomi Enterprise Search