July 12, 2007

Everything Is Miscellaneous

by David Weinberger

I'm sorry, but no. Everything is not miscellaneous. This intriguing, lively, and useful book is also, I believe, mistaken.

Weinberger seizes a good and interesting idea at the outset. Organizing physical objects — arranging books in a library — is very difficult because each book can only be in one place. Arranging a physical index like a card catalog is easier than arranging a bookcase, because cross-references let us build alternative hierarchies of subject, author, title. Better yet, we can skip the hierarchy and jump to a faceted classification system in which each axis of classification is independent and coequal.

Besides, it's easier to move physical proxies like notecards than to move the objects themselves. But building an index to an index — a database or a Web — gives us even more flexibility and makes reorganization much easier. Because the database is virtual, we need not limit cross references for fear that we'll run out of drawers.

First year computer science majors know all this, and knowingly repeat to each other: 'the answer to all problems is another layer of indirection.'

This book, especially its early chapters, is a useful corrective for some of the more exuberant fantasies of information architecture and for some naive interpretations of the Semantic Web. Weinberger sets off at interesting place and commences with a valuable contribution. Multiple organizations are better than a single, bad organization. Simple search is better than a bad organization. Letting people muddle through, discovering whatever ad hoc structure they can manage to find, is better than trying to teach them all an arbitrary and peculiar scheme.

But does this mean that everything is miscellaneous? Or does it simply mean that a miscellaneous structure is better than a bad one? A math student, facing a problem set that seems baffling, might conclude that "everything is too hard", but that student's frustration does not prove that mathematics is vain.

Everything is Miscellaneous is so thoroughly popularized and simplified that subtle and perplexing questions — where does meaning really reside in a hypertext? can formal notations usefully represent knowledge? what are categories, anyway? — are described and dismissed in grade school vocabulary. We are given a bare minimum of references, and most of those defend the facts in the author's mundane examples without giving much insight into the extensive debates and controversies that underpin the ideas.

For example, RDF triples are introduced with a vague wave of the hand, and dismissed with an even vaguer allusion to the advantages of fuzzy sets. Microformats appear briefly, are praised for miscellaneousness, and vanish. Conceptual clustering is trotted out to rescue our metadata from arbitrary categories, is lauded for being open in some way to data mining in the tagosphere, and nothing more is heard of it. The treatment of hypertextual structure and the construction of textual meaning is trivial, and there's no hint in the text of the rich discussions and violent disagreements it has inspired. We seldom see real examples or discuss actual methods; it's all "Capri, Italy and capri pants."

In the end, the conclusion — “Everything is miscellaneous” — seems calculated to please managers who are too lazy or too distracted to actually master their business. If everything is miscellaneous, what would mastery mean? Instead of coming to conclusions, we can assure managers that no conclusion is possible, that the only wise course is to throw everything into a heap and let the customers sort it out.

That might work for retail. I'm a peddler, too. But there's a big part of the world where the right and wrong answers are not simply what today's customer happens to imagine they believe, and in that part of the world, I think, everything is not really miscellaneous. When your control loop is unstable and your distillation column runs away, that's not miscellaneous. When your job shop scheduling breaks down so your people are changing paint colors every five minutes and your customers can't get the machines they ordered, that's not miscellaneous. Some things are arbitrary. Some things are matters of linguistic or social convention. That doesn't mean everything is arbitrary or conventional; "everything is intertwingled" means that the structure is complicated, not that it doesn't (sometimes) exist.

Of course, Weinberger knows this. He's a philosopher; these controversies and perplexities are where he lives. But here it's all simplified and popularized.

Some things are related, other are not. It's all hard to understand, but the height of the mountain does not make the climb inconceivable.