Micro-blogging models

There are many ways of sharing small snippets of interesting or useful information, usually for a) one’s own recall; b) interested friends; and c) Google fodder, if you have some genuinely useful comment to add. I would like to briefly mention three models.

The closed model

This is typified by Facebook. Wall posts, status updates, and so on, are used to associate small pieces of data to different objects such as people, events, and items (using Open Graph™). The information architecture is so comically terrible and shockingly powerful that I recoil in horror. Everything that goes in gets locked in to Facebook’s jumbled taxonomies. The ontology is no more advanced than identifying things by a Facebook-assigned identifier.

The worst thing about the model is that it is essentially impossible to re-use anything in it. Today, I saw something funny one of my friends had posted on Facebook, wanted to re-publish it (‘re-tweet’ in Twitter terminology; ‘like’ in Facebook lingo), but I could not. The only way to like an item on Facebook is to press the thumbs-up ‘Like’ button and let Facebook manage everything. There is no way for a user to get a permalink, no tag or identifier for referring to the object outside Facebook, and certainly Facebook never allows anything friendly like an Atom feed to free anything for re-use.

I hate this model, vehemently. It fails my most basic expectations and tests.

Open models

Unstructured

Twitter is the prime example. They are very good at this, using standards decently, and solid Atom support. Twitter is in fact the ideal Social Web application (grunky terminology).

My main problem with Twitter is that its ontology is just a little too simple; the information architecture is sound but too minimal. The real practical implication of this is that almost everything posted there is inconsequential noise (or worse, junk). The model can be used to produce genuinely worthwhile links and updates, but is usually not. The only implication for me is to give it a miss.

Structured models

I would like to promote the under-dog here: Google Reader. Without much of a following that I can see, Google has done almost everything right that Facebook has messed up. Picassa, Profiles, and finally Reader are simply superb. The great thing about the way they work together is that they acknowledge the scope of social networking tools. Facebook goes badly wrong when it tries to start owning content, and Twitter when it encourages tweets themselves to be the content. Google on the other hand has never been interested in owning content, but has expanded from its beginnings indexing it for search to providing a framework for linking and meshing it together. Reader is the key to this.

The thinking goes that everything worth following is in a feed somewhere (preferably Atom with some RDF-transformable metadata dependent on which camp you sit in). Reader takes feed items, and knits together the content your produce and read, letting you share things (like retweet) with comment, and lets you add your notes or micro-posts too. Unlike the Twitter model, which lets you contentfully retweet but encourages junk, Reader by default links together richer entries using Atom, not just naked URLs. The end result is that the average entry ought to be much more worth reading than Twitter.

So, I am still exploring the social web ideologies, but like Google’s ecosystem lots and lots at the moment. The big thing really holding it back so far as I can tell is the strong lack of penetration of any particular metadata language. Reader has to put up with a lot of RSS junk, malformed Atom, and any metadata embedded comes in too many forms to make any good use of.

Any gripes? The usual: their tools for generating mashups use document.write, breaking things for XML sites like mine. HTML5 will fix that.