Revisiting xFolk

The key to creating a successful microformat is to keep it simple and pointed at specific problems. It should also build on previous efforts to help gain acceptance. To this end, it's useful to revisit the exact problems xFolk is intended solve and then examine other current relevant efforts.

What exactly is xFolk supposed to do?

xFolk is designed to allow people to publish tagged links in web pages using standard xhtml mark-up. The primary purpose of these tagged links is to enable their aggregation, typically by search engine and other crawlers that automatically extract the information from the pages using xFolk markup as their guide. These crawlers may exist on the Internet at large or be agents such as search toolbars on the user's machine. Specific instances where xFolk could be used are:

• Publishing a link blog similar to Kottke's remaindered links.
• Publishing a link blog from an archive maintained by del.icio.us or a del.icio.us clone like scuttle (a working instance of which has graciously been provided by Niall Kennedy).
• Tagging a link in a web article that you want to indicate as related to a whole set of other artifacts that the reader could then access. This type of tagging indicates how you are thinking of the link as you use it, as opposed to how you think of archiving it. Tags added to links in this way could simply be appended to the link by any sort of aggregation agent.
• Indicating how you would tag a particular article that you author.

This scope overlaps somewhat with Jeremy Zawodny's call last Fall for the development of a personal del.icio.us or Danny O'Brien's ideas for Macintosh desktop search features. However, from what I have described of xFolk, it should be clear that it is much less ambitious. xFolk is just a modest attempt to create a standard for embedding tagged links in web pages so that they can be shared with the outside world. However, the third bullet indicates one slightly ambitious feature of xFolk, also hinted at by Zawodny and O'Brien: allowing users to include as much information as they can about the link, when they want to add it, not just when they are specifically turned to the task of archiving items.

Existing format efforts that share goals with xFolk

There are a number of existing format efforts that share goals with xFolk. For the most part, these formats specialize in publishing collections of tagged and commented links. They include:

• Attention.xml which gains most of its expressive power from the XOXO microformat with its particular emphasis on lists. It also references the technorati mechanism for tagging posts.
• Scuttle, the open-source del.icio.us clone referenced above, that publishes its links in a definition-list xhtml format with classes defining semantics of the different components. In design, it seems to be inspired by some elements of Attention.xml.
• del.icio.us, itself, which exports links in a format using hypertext <a> elements and <div> elements with classes to define the semantics of different elements. Joshua Porter provides one example of how this might be used on his blog Bokardo.
• Technorati's tagging mechanism for posts and which relies purely on the hypertext <a> tag with the rel attribute set to “tag”.
• Self-publishing efforts such as Jason Kottke's remaindered links which uses a list but with no effort to define any semantics. Kottke is particularly notable because he now makes his living by blogging and remaindered links has been cited as a resource by mainstream media.

In examining this sampling of formats, a few things stand out. Those that are machine-generated (i.e., attention.xml, scuttle, and del.icio.us) tend to have richer semantics. By contrast, formats that are generated by people, like Technorati tags and Jason Kottke's remaindered links, tend to have much lighter semantics. Further, formats that are oriented toward publishing aggregations of links (i.e., attention.xml, scuttle, del.icio.us, and remaindered links) tend to use a list of some sort but differ in type of list.

It seems no accident that the machine-generated efforts are, to a one, oriented toward publishing lists of links. After all, that is the type of mass-production best-suited to a machine. Further, it is not too surprising that machine generated efforts tend to provide more semantic markup, particularly those that are open source. The designers of those systems are more aware of the underlying semantics to begin with because they have had to create the database schemas to hold the data. Further, open source developers tend to be more interested in making sharing easy, hence the focus on semantically labeling the html produced.

In this context, Kottke is an interesting exception. He publishes his list of remaindered links using Movable Type, a blogging platform. Blogging platforms are generalized web publishing frameworks that do not contain specialized schema for storing things like folksonomy tag data. Rather they contain general data fields, different by platform, that with some effort can be turned to the purpose of narratively publishing links with commentary and tags, i.e., exactly what Kottke is doing. Inserting tagged links in a blogging environment typically requires hand-coding of the relevant values for each entry, hence technorati's strategy of providing a format that is easy to code by hand.

Synthesis

A little reflection leads one to conclude that, in general, the Kottke case is more likely the rule than the exception. Setting oneself up for blogging can take as little as five minutes leading to a doubling of the number of blogs every five months for the last two years. In contrast, figuring out how to use del.icio.us requires significantly more work, and setting up a del.icio.us clone, like scuttle, requires access to a server and the willingness and skill to resolve installation pecadilloes.

Given all of these observations, I'm led to three salient conclusions regarding how current format efforts should influence xFolk:

• The type of semantic markup used in attention.xml or scuttle likely cannot be transferred wholesale to linkblogs like Kottke's remaindered links. In an important sense, the underlying data schemas supporting blogging platforms are just not rich enough to support easy and systematic translation into the structures required by attention.xml or scuttle. A simplification of these structures seems in order for human use.
• The first point notwithstanding, the semantic markup used in attention.xml or scuttle does not supply the required flexibility for easy personal expression required in personal publishing platforms like blogs. For instance, I should not be forced to use a list format to express tagging information in the flow of a paragraph.
• Semantic markup from efforts like del.icio.us, attention.xml, and scuttle may be used as a basis for constraining and informing the development of xFolk. One might imagine that there could be a heavy version of xFolk that would be close to the list models used by attention.xml and scuttle and light version of xFolk using things like <span> elements for more prosaic use in the middle of paragraphs, etc.

Next Steps

This post was intended as a set up for a revision of xFolk iteration 2, published to this blog on Thursday. Having spent time synthesizing feedback from various parties (including Tantek Çelik, Kevin Marks, Thomas Vander Wal, and Josh Porter) on that effort, I've come to a better understanding of what xFolk should do and how it should do it.

As always, I appreciate feedback via email, comment trackback, or technorati tagback (you can do tagback by copying this link into your blog post and pinging technorati).

Technorati Tags: xFolk