C[omp]ute

From: "andrew cooke" <andrew@...>

Date: Tue, 19 Dec 2006 11:15:57 -0300 (CLST)

http://radar.oreilly.com/archives/2006/12/google_depreciates_SOAP_API.html

They're switching to the "AJAX API".
http://code.google.com/apis/ajaxsearch/documentation/

That means that the API you program against is defined in Javascript
rather than as a messaging protocol.  This is a neat approach - you
isolate the client from the implementation (in theory they could implement
the AJAX API with SOAP, but presumably they use GET/REST and return direct
HTML).  It's similar to what we do in NSA with the "client" instance of a
service.

One of the advantages of Java 1.6 is the close integration of Javascript,
so Java could call this directly.

A disadvantage of this approach is that, since it's an AJAX interface, I
assume that the return values are HTML.  If you want to process the data,
rather than present it, that means parsing the result (DOM or XSLT) - and
presumably the result is not in a fixed, well-documented format.  But from
Google's POV this is probably a feature - they're reducing the likelihood
that a middle party will subvert their branding/presentation/profits ("do
no evil" having long ago shifted to "do nothing to hurt the stock price").

Another possible problem is that persistence seems to be in the Javascript
layer.  In some ways that makes sense.  How would Google provide
server-side support persistence for third parties?  Caching can be
provided at the client (transparently, as far as the API user is
concerned), reducing server load.  However, experience with AJAX on the
Portal has shown that client-side state can be a problem when you want to
change pages.

More broadly - is this the sign that SOAP has failed?  No-one has any
faith in auto-generating code from WSDLs.  Will the VO catch on?

Andrew