Laughing Meme

K: What was that thing we used to talk about before “tagging”?

A: Um… keywords?

K: No, fancier

A: Topic maps?

… searching ensues …

K: Faceted classification!

Some tags I check when trying to figure out who someone is/what their story is: me, friends, work, home, weather, craigslist. Also a quick visual scan for place names.

Even folks who’ve managed to stay fairly anonymous leak a lot of info in their tags.

Lots has now been written on schemas for storing tags (mostly in relational databases). In fact Tag Schema is a blog and mailing list devoted entirely to exploring that space further. (maintained by Nitin Borwankar, working at Odeo these days).

So what happens when we take off our DBA hat, and put back on our programmer hats? (because really, who has seperate DBAs these days?)

Most implementations I’ve seen treat tags as something you put in a bucket attached to an object. (could be most implementations I’ve seen are in Rails which encourages this)

But tags aren’t something an object has, their something an actor said about an object. An actor is key. Tags without someone/something behind them are devoid of context and meaningless. So rather then coding it as:

object.tags << sometags

consider

user.tags(object, sometags)

where tags is no longer a noun (a collection), but a verb.

At some point I don’t care how you store it in your database, but you need to be thinking about it as something a user is doing to the object.

For the folks I promised an article on how to implements tags in SQL. Sorry, I never finished it. I got sucked into some more esoteric problems, and never got back to writing down the basics. However, all is not lost. Peter has done a very nice job of writing about how he implemented tagging for his snippets code libraries.

1. Matching on data intersection across a many-to-many join – aka, how to find an object (e.g. a post or a recipe) with two or more tags (i.e. show my all posts tagged with tagging and sql). For the lazy, skip to bottom for a solution. (but it’s worth reading through) This is what enables tag combos. (Also available from snippets)

2. Find items with similar (or as many as possible) relationships – for a ‘related posts’ box etc – aka, find other recipes tagged with similar tags to the recipe I’m currently looking at I’m currently looking at. In Recipes on Rails, if I’m looking at my recipe for hot chocolate (tags: beverage, hot, chocolate), use this query to also show tea and coffee. (both tagged beverage and hot)

3. Find all many-to-many relationships which are tied to an arbitrary number of other many-to-many relationships – I’m looking at all recipes with the tags easy, and hot, what are related tags I could use? How about carrot, and curry (my curry carrot soup is tagged easy, soup, carrot, and curry, not to mention winter and favorite), or maybe silly with which I’ve tagged my boiled water recipe.
   After some futzing I came up with the same query that Peter did, but I was really hoping there was another solution, as I’ve been seeing some of the same worrying numbers using MySQL 4.1.x’s subqueries that Kevin is.

There you go, you no longer have any excuse for building a sub-par tagging system.

fyi, Peter, from whom all those tips originate, on top of being the creator of the snippets site, is available for hire.

People are still too stiff and rigid with their tagging technique. Loosen up. You don’t have to find the “right category” to put something into, that is part of the tyranny and inflexibility of a classification scheme that we’re trying to get away from. Don’t tell me what it is, the “truth” of it as it were. Tell my why it matters.

For example I use the tag “inspiration” to keep track of ideas I want to steal, or think about more on my various projects. (inspiration+redesign are my first notes towards a Magpie re-design)

Variations on toread, and *toread are in wide use as useful meta-tags, and a handful of people are using variants to track specific research projects, or tasks.

Marnie is experimenting with the tag nptech to build a community of non-profit tech workers, we’re using a different tag for the our anarchist tech work, and there are a handful of bibliographies being organized around a specific indicator tag.

And lastly don’t be afraid to build combos. One of the key ways tags work is the set logic of multiple tags. They’re your links on redesign inspiration, javascript usability, Boston bakeries, or Photoshop books. Much like with wikis, the meaning arises not from the individual components, but when you ram them together to indicate a single new concept. (and I won’t even make you use CamelCase)

Technorati Tags is the latest tag implementation, mixing blog content into the slurry of tagged links and photos. You can either use some special snippets of HTML to signify tag membership to their crawler, or you use the “category system” of your RSS/Atom feed.

While they don’t go into great detail about which “category system” they support, we can hope they support looking in dc:subject, one of the few elements that is explicitly repeatable in an RSS 1.0 item. (and if they don’t, I hope they take this as a hint. Additionally I really wish del.icio.us RSS feeds would stop wodging all the tags for a given link into a single dc:subject, but instead go the multiple dc:subject route)

Making it Work w/ MT

Anyway, that isn’t what I sat down to write about, plenty of other people have chronicled Technorati’s tags sufficiently. I just wanted to offer up a trivial little MT plugin that I use to split the keywords field into a list, thereby making the field suitable for use as a low rent tag system. I do this with my MLP feed to simulate tagging for import into del.icio.us (being contrary, I go the opposite direction of everyone else, and publish my linkblog locally, and then submit it to del.icio.us)

The plugin provides a container tag MTKeyWordList, and a content tag MTKeyWord, that I use like so in my RSS template:

<item>
...
 <MTKeyWordList>
    <dc:subject><$MTKeyWord$></dc:subject>
  </MTKeyWordList>
...
</item>

Alternately if you wanted to use keywords to add the custom Technorati HTML you could do it like so

<MTKeyWordList>
 <a href="http://technorati.com/tag/<$MTKeyWord$>" rel="tag"><$MTKeyWord$></a>
</MTKeyWordList>

You can grab the keyword list plugin here.

Btw. I really wanted to use the following link to follow TechTags discussion, but unfortunately it doesn’t actually work, maybe some day:

http://del.icio.us/tag/technorati+tags|tag|folksonomy

update: alternately George’s TechnoratiTags plugin works just as well, though I personally prefer the flexibility of my approach.