I agree with you that the Semantic Web (as defined by the W3C) is fighting an uphill battle (but not necessarily doomed to failure). That’s why I’m a big advocate of a more iterative approach towards the semantic web (lower-case). Specifically, I’m very interested in Microformats for several reasons. First, Microformats start by encouraging best practices that web designers and web developers should already be following (plain old semantic HTML) and then add more semantic data from there. Second, Microformats are designed primarily for humans and secondarily for machines (no invisible metadata). Third, Microformats are based on existing open standards (the “pave the cowpaths approach”). Sure, the scope for Microformats is much smaller than the W3C’s Semantic Web (e.g. it’s not an extensible vocabulary) but this less ambitious scope is probably a good thing. I think the W3C’s Semantic Web will eventually succeed but I think that other, simpler, semantic web technologies (like Microformats) will become successful first.

I hear what you’re saying – there is a lot on the Internet that works better when we structure our data in a meaningful way (i.e., semantically). However, isn’t the problem that Tim Berners-Lee originally created the web as a visual medium? In other words, HTML exists not to structure data so that computers can understand it (semantically), but to present data to people (visually).

As I see it, web pages started out life as pure presentation layer (therefore visual), but we have piggy-backed on things like meta-data for search. While we can do much more with the web when we structure it well, it wasn’t originally intended to work this way, and that’s why we have problems – because what we have is a hybrid model.

Now, there are two approaches to solving this problem. One is to introduce a new model for the web – one that provides both a data model (for machines to read) and a view (for people to read). (We already have controllers – web browsers.) This approach is the one the “semantic web” proponents are working on.

Personally, I think this approach is doomed to failure, right along IPv6 and a whole host of “better” techologies. The problem is that every day that it takes to develop these new approaches is a day that the old standards become more entrenched – and frankly, better able to accomplish the same results. From this perspective, I think you are correct. We do need to move more toward intelligent structuring of data such the presentation logic is separated out. But the transition is going to be painful, because the original design didn’t take this thinking into account.