[OpenSpending] messages about standards

Fri Oct 5 19:54:35 UTC 2012

Hi James,

On Fri, Oct 5, 2012 at 5:20 PM, James McKinney <james at opennorth.ca> wrote:
> On 2012-10-05, at 10:08 AM, Friedrich Lindenberg wrote:
>
> As for the others: XML and CSV seem very realistic for me, both have
> great tooling and are widely used. Linked Data is something that I
> have a personal problem with. It would be interesting to survey other
> data consumers, it seems to me like LD is mostly a proposition made by
> data publishers with fewer people actually wanting to use it.
>
>
> Could you expand on your problems with linked data (by which we mean RDF, I
> assume)?

I'd prefer to avoid this: while I love a good RDF flame war it's
likely to derail this debate. Still, if you know data end-users
(developers, analysts, journalists) who would really want an RDF file
over JSON/XML/CSV to get their job done, I'd be very keen to learn
about them and the way they do things.

> In general, there should be a preference to publishing the same data in a
> variety of formats. For geospatial, I know many governments publish in both
> shapefile and KML. Shapefile is favored by the GIS crowd, whereas KML is
> often the format sought by web developers. CSV is the most accessible
> structured format. I wouldn't put (the XML serialization of) RDF that far
> behind XML in terms of complexity.

Don't get me wrong: I don't want to stop anyone from using their
favorite file format.

But to me it feels like a good standard is something that can actually
be consumed by a reasonably small application. In fact, the whole
point of standardising something seems to me about reducing options
(and therefore complexity). If you allow data to both be expressed in
formats A and B, you end up with two options: either you allow people
to publish one format and not the other - then your consumer has to
support both - or you mandate that both formats must always be
produced - then your producer is going to be more complex than
necessary.

All of this comes, in my opinion, on top of one fairly simple
observation: that there is really nothing wrong with CSV for
transactional data. Sure, it isn't typed and people often don't really
do CSV when they say CSV - but we can fix typing in the spec and for
CSV parsing there is just a truckload of great tooling. The fact that
non-techies can open it in one of the four computer programs they
vaguely know how to use just adds extra sweetness to this.

So: sorry about the RDF flame-bait and keen to hear your opinions :)

Cheers,

- Friedrich