It’s all very cool, but I find myself wondering about the implications of a library where users can’t browse the collection. At my own library, we have a research collection whose stacks are closed to the public. While this means that material can be preserved – there are some extremely rare, valuable and fragile works in the collection – the opportunity cost of such a setup means that patrons can’t browse the collection and discover things they didn’t know they wanted.

Have you ever been looking for something in a library (or bookstore) for something and come across a book on your topic of interest that you didn’t know about? A book, perhaps, that you only came across because you stopped focusing so narrowly on a small section and let your eyes wander over the entire length of a shelf or height of a bay and that your eyes seemed to lock onto almost automatically? Well, in librarianship, that’s called serendipity, though I’ve seen it referred to in retail circles as “browsability” or “walkability”.

Serendipity has always been a big part of a library’s appeal. It’s one of the (many) reasons things are so carefully catalogued: in addition to providing organizing principles by which material can be sorted, it also allows patrons to discover interesting titles simply by walking through the aisles.

See, people often have a good idea of what they’re looking for, but the human brain is especially adept at making connections and to rapidly sift and sort data (e.g., a shelf or bay of books), turning it into information (e.g., a particular title that relates to a patron’s interests.) Our brains naturally look for relevant “hits” – that is, things that seem to stand out – and when presented with a large dataset, will tend to automatically narrow and widen our focus to try and make sense of it.

Serendipity, in that sense, is about trusting that the patron is allowing their brain to naturally narrow and widen its focus in their search.

This serendipity, derived from a collection’s browsability, is largely lost online. Computers are very good at storing and sorting information, but they are extremely poor at making the kinds of connections between things that come so naturally to the human brain. It’s why companies like Amazon spend billions of dollars trying to mimic a highly complex and nuanced process that, for the human brain, often seems effortless.

There are all sorts of computer models for mimicking these connections. The most popular is probably purchasing habits, which you’ll often see online as “people who bought this item also bought…” or some such similar phrase. Another is by collecting and comparing attributes (sometimes hundreds or even thousands). The Music Genome Project is a good example of this. If you haven’t tried Pandora, the commercial Internet radio derived from the Music Genome Project, you really should.

But crucially, these computer models only mimic human behavior and often poorly at that. I’ll bet you’ve gotten frustrated with at least one online store that offered “recommendations” that had little or nothing to do with what you wanted. Assuming that the company isn’t putting in bad data, this is usually because the system is trying to generalize all shoppers’ habits; however, individuals often come to different conclusions about the same set of data, because due to many factors (e.g., life experiences, political opinions, race, gender… even how recently they’ve eaten can have and effect!), their brains will automatically assign different mental “weight” to data.

We’re getting better at training computers to deliver quality recommendations, but what we’re discovering is that the computers have to keep a lot more data on people in order to deliver quality results (which naturally leads to privacy implications). Amazon, for example, has several enormous datacenters around the world and has to store, sort and index terabytes, petabytes, even exabytes of data… just to make sure it’s recommending things you might actually like.

The only way to make up for the lack of a browsable collection is to put resources into creating an online catalog that will accurately mimic the abilities that the human brain innately possesses. Recommendations through “social” technology, though still in their infancy, seem to hint at a possible solution, but as yet it’s unclear exactly how “social” fits into providing better results (especially without running afoul of privacy concerns).

I don’t expect that all libraries in twenty years will look like the Mansueto Library. But given the costs involved – building an automated storage facility, designing a truly intuitive catalog that consistently makes top-quality recommendations – wouldn’t having librarians and pages to catalog and shelve the books in open stacks cost less and still allow for browsability and serendipity? Or are librarians destined to react to such automation as Ewan MacGregor’s character did in the first twenty seconds of this clip from Danny Boyle’s (highly underrated, in my opinion) A Life Less Ordinary?

The worst part of it all is that “www.” was never really necessary in the first place. Back in the early- to mid-1990s, when the Web was just starting to take off, we were all using relatively simple web browsers. Even so, I remember that the first browser I used heavily, Netscape 1.1, would automatically fill in the “http://” protocol if you didn’t specify it. That was an obvious call, because hey, you were using a web browser, so the odds were pretty good that you were going to a web site if you didn’t tell the browser otherwise, so the of course the browser should go ahead and fill in the protocol.

The “www.” part of a web address is actually a subdomain of the main domain, which was supposed to help separate the web server (the part of the server accessible via HTTP) from other protocols, like Gopher, FTP, email and the like. However, if you tried to connect to servers using HTTP, the servers were almost entirely capable of noticing what protocol or port was being used and answering correctly. Unfortunately, no one at the time had any idea that the Web would become one of the most transcendent ways of sharing information, ever, so sysadmins everywhere just gave the Web server a subdomain (e.g., www.somesite.com) as they did with, say, mail servers (e.g., mail.somesite.com).

Fast forward fifteen years and the Web is not only still around, but it’s going strong and still growing…which is more than anyone can say for poor old Gopher. However, we’ve been stuck with “www.” add-on, even though web servers are smart enough to perform a simple forward if it’s been set up.

It’s time to drop the “www.” from our domains. Don’t worry, though; your sysadmin can set up a simple forward so that if people still type in the “www.”, they’ll just get shown to the right URL.

There are several reasons I’m advocating this:

1. There’s no need for the “www.” part. As I said above, it’s completely vestigial in terms of the Internet. There’s virtually no chance any more that someone’s coming to your site with Gopher.
2. Less is more. Leaving off the “www.” gives you four characters back and these days—with social networking sites like Twitter being as popular as they are—the shorter your URL is, the easier it is for people to remember it and pass it along.
3. It looks better. Okay, this is a personal reason, but I happen to think that a shorter, more readable URL looks a lot better on an advertisement than one with the “www.” included. Whenever I see an ad with it still included, I feel like the company behind it doesn’t quite get the Internet. (Embarassingly, when Chicago Public Library started their “Not What You Think” ad campaign and had ads all over the city referring people to “chipublib.org”, they discovered that their web server was not set up to handle automatic redirects from chipublib.org to www.chipublib.org.)
4. It’ll do better in search engines. Google, for example, ranks http://somesite.com/ slightly higher than http://www.somesite.com/.

So brush up on your .htaccess terminology and drop the useless “www.” from your domains.