Raw

This may seem a strange request from an RDF fan who is on record calling OPML bloody awful (or words to that effect). But I'd like to humbly ask that anyone exposing RDF services on the web consider also emitting the following :

1. Content supplied as RSS 2.0 and Atom
2. Simple resource relations as OPML

Both of these should be made available through a simple HTTP GET. This is looking towards what I've been calling Out of Eden (warning: that's a search on the old site data, things are still in transition). Make SemWeb data available for viewing through RSS readers and the like. Give the innocents a bit of knowledge. Hey, they've got the coolest UIs, they deserve some Earthly Delights. (But don't make the mistake of equating innocence with stupidity, that leads to bad Karma).

There are a lot of intelligent and capable developers out there working on the web around syndication technologies, using RSS 2.0 and OPML not for any technical reasons (they would have more sense) but because they are de facto standards. There are applications using this stuff which are considerably more visible than the Semantic Web apps out there. But like pretty much any interesting software system, the benefits of Semantic Web technologies need to be seen to be believed.

I'm prompted to post this now after seeing Grazr (spotted by Adam), which is a neat little Javascript data browser, not unlike the column-based filesystem browser on the Mac. I'm sure given a bit of effort the same kind of thing could be built to operate directly on RDF sources. But it'll be a lot easier to serialise out (maybe through SPARQL+XSLT) to formats that existing widgets and tools understand.

RSS 2.0 is lousy, in particular with the broken-as-designed content escaping frozen in the spec. OPML is bloody awful, in particular in pretty much every respect. But data is easy to output in these quasi-XML formats, just don't expect to be able to read it predictably. This isn't really an issue for most display purposes, and RSS and OPML tools tend to be liberal about what they consume (of necessity because of the poor specs). It doesn't really matter exactly how you flatten RDF down to these formats, though some expressions are obvious - e.g. if there's a dc:title and dc:description, wrap them up as title and descriptionin a no-namespace RSS item. Where URIs are dereferenceable, property relationships can be squashed down to a tree of links in OPML.

Ok, so this in part is trying to appeal to people who wouldn't otherwise give RDF a second glance. I personally think RSS 2.0 has been legacy since long before RFC 4287 came along. OPML is a badly-drawn toy format that's riding on the back of RSS. So why aren't I suggesting more capable formats (microformats, direct RDF serializations...)?

Because the simple-is-good message has been louder than good-is-good. There are a lot of tools around that already support RSS 2.0 and OPML. These formats, shored up with HTML and HTTP, are more than adequate (assuming liberal interpretation) for fairly sophisticated display purposes, as is being demonstrated all over the place. For anything else there are serious issues, RSS 2.0 and OPML are non-starters when it comes to shifting data around.

Why aren't I suggesting more capable formats? Because I don't need to. The web's natural selection will see RSS 2.0 and OPML fade as developers choose formats that can support the functionality they want, without being arbitrarily crippled by bad specs. But right now there's a great opportunity to make existing apps (aggregators, OPML browsers) that much more useful by providing them with more interesting data. So what if you-know-who decides to patent/lock down OPML? Ten minutes work swapping to a different serialization syntax.

Five or so years ago it was really cool what you could do with XML-RPC, and its existence got quite a lot of people to think about web services. It was a blind alley in itself (it did lead to SOAP, but I'd argue that SOAP's place of utility isn't really the open web). Whether or not progress would have been faster without it is another question. But many of the early adopters of XML-RPC were the first to drop it when better alternatives became apparent (such as RESTful XML doc/literal). So I'm suggesting the best way to get past rubbish like OPML is to acknowledge people are doing good things with it, and engage with their tools in their terms. They will be curious about what's on the other end of the wire, and if the intuitions surrounding Web as Platform and Web of Data are correct, the SemWeb comes faster.

In the meantime, RDF-heads get some cool UIs at low cost.

@en