[ckan-dev] Steve Yegge on Platforms

Mon Oct 17 13:13:08 UTC 2011

Dear David,

On Mon, Oct 17, 2011 at 11:18 AM, David Read <david.read at okfn.org> wrote:
> Really interesting article - cheers for that. I'd agree it's good to have a
> good API, although I'd say that Google have been pretty successful
> generally, despite their poor APIs ;-)

Yes, but the question is: how sustainable are sites like G+ when they
fail to enable external innovation? This is kind of our key strength:
there isn't even a strong distinction between the inside and the
outside, so it can naturally happen.

> I think you're being a little unfair on CKAN though, because the API has
> given full access to the data since version 0.6, over three years ago. Sure,
> we have had a few private services that manipulate that same data, such as
> tag completion, but since these are trivial operations I don't see it as
> much of a barrier. And we've also built a permissions model, which has not
> had explicity APIs, but that is set to change:
> Since the Spring we've been working to make channel our Web UI's
> interactions with CKAN data via a new 'Logic Layer'. This is also exposed
> *in full* to the world as the 'Action API'. This API has not been announced
> yet, but a trial version went out last month in CKAN v1.4.3. Check out the
> docs: http://readthedocs.org/docs/ckan/en/latest/apiv3.html It's designed to
> be friendly to calling with javascript, so it's standard RPC, not onerous
> and provides good status and error messages. It is also designed to return
> help text (although there's not much there yet) and we have planned for it
> to be discoverable too. So once we've got that, I think Steve Yegge would
> approve.

I didn't mean to imply that we had no API or that the API was bad (I
do think we have a few long-standing kinks we need to address, but it
has repeatedly proven very useful). Its just that visiting
thedatahub.org, you'd have a hard time noticing it: the link is hidden
in the footer and points to a somewhat chaotic wiki page which in turn
only points to the older version of the API (I hadn't known the v3
thing had docs, so that's my fault).

Maybe we could include ckanext-wordpresser on the main thedatahub.org
site and link to some of the cool things that have already been done
with the API, e.g. the LOD2 cloud, a writeup of harvesting, the UK25k
collector, and the visualizations someone did for publicdata.eu.
Particularly as the API starts to extend beyond metadata into data,
this should become an impressive showcase and I'd love to contribute
what I have (in terms of scripts running against webstore).

We just need to make sure that we don't get all the cool features but
nobody who's not on ckan-discuss knows about them.

- Friedrich