"Publishing Open Data - Do you really need an API?"
Jury Konga
http://www.peterkrantz.com/2012/publishing-open-data-api-design/ I
know this group will have some opinions! Look forward to seeing the
feedback.
Cheers Jury
Ryan McGreal, Raise the Hammer
In a sensible web service API, the client will retrieve a given data set
using an HTTP GET request on the URL for that resource. That's how the web
works, after all - and if your web service architecture requires clients
to, for example, push big fat XML payloads across the network using POST
requests against a SOAP endpoint, your real problem is that you're trying
to serve RPC over HTTP instead of just using HTTP.
The author suggests that you should publish a "data dump" to your web
server, presumably by dropping a static comma- or tab-delimited text file
to a static folder exposed to the internet. However, the client still has
to do an HTTP GET request on the URL for that data.
If requests for a static text file are fast and responsive but requests
for data from an API are slow and cause the server to bog down, that
suggests an issue of slow, blocking database reads (missing table index?),
inefficient request processing by your web application, poor response
caching, etc.
At a more basic level, the author's approach suggests that the API is
being added to your web application as an afterthought. A well designed
web application is built on an API in the first place. In that case,
making the API available to third parties becomes a relatively simple
matter of exposing it, rather than bolting on a parallel application that
serves structured data rather than HTML.
Regards,
Ryan
