You're right, but.. I cannot help but think about that ridiculous orange xml/rss icon that seems to be ubiquitous these days on web pages. Autodiscovery is so passé.

Having both human-readable and machine-readable information is ok, I think. If you want to put the icon then do - it helps to distinguish blog posts that are reviews when scanning through a page. I don't think there should be a requirement for the icon to link anywhere though, and there needs to be extra metadata so that a machine can understand that you're reviewing a paper, and which one.

What ever happened to TrackBack-enabled remote comments on HubMed?

I turned the TrackBack off in HubMed - the only people really using it were spammers.

Good points. Obviously the icon isn't going to appeal to everyone, and we're going to offer an icon-free solution in a few weeks.

Postgenomic is great as far as it goes, but it has the limitation that it doesn't distinguish between thoughtful analysis and simple links to abstracts.

The extra metadata is coming, too, Alf. Give us a little time -- we're only a couple months old!

Dave: if people used rev="review" then Postgenomic would be able to distinguish between serious reviews and random links, so that's not really a problem.

I'm glad that you're encouraging people to add semantics to their posts anyway, I'm sure there'll be a good solution in the end.

We're working with Postgenomic to ensure we're not duplicating their efforts. The two sites have complementary purposes. We're requiring a little more effort on the front end, which we think will result in a more usable system for many readers.

One thing we could do is automate our system to add the rev="review" attribute to posts on BPR3, which means they'd be indexed in both systems.

That would be good if you could automate the 'rev' attribute.

Maybe add a 'rel' attribute as well, that would indicate that the person writing the review considered the thing at the other end of the link to be a 'serious paper', whether it was published on a blog, in an institutional repository, or wherever?

particularly frustrating from my point of view, that of a author and not a tool developer, is that there is no easily accomplished and standard way to say, "Here's the info on a paper(s) I'm talking about, and here's the info on the post talking about the paper."

Postgenomic recommends that rev='review' only be used when your post is about one paper in particular. hCitation has a means for dealing with multiple reviews per post, but it requires you to stick hard-to-remember classnames in various html elements where they don't really seem appropriate(like why is the summary in H2?) It's really more suited for simple stuff, like here's a site talking about X and I give it a rating Y.

Perhaps I don't fully understand it, but it seems fairly complicated to implement if you'd like to include title, journal, date, authors, volume, page, DOI, PMID, abstract, and your summary all as separate pieces of data, which it seems like is what you have to do if you want to be able to do something useful with the aggregated reviews.

What I'd like to be able to do, for example, is say "Of the papers being reviewed on weblogs, how many papers for keyword X were published in which journals, grouped by year?"

Or to compare the frequency of article reviews among journals, and group them by which are open access and which aren't.

Does that make my frustration clear? I'd like to contribute metadata, but it seems unnecessarily complicated for me to contribute the atomized kind of info that would be really useful.

I don't see why you couldn't use rev="review" more than once within a post. With that, an aggregator could do all of the things you mention: the aggregator can use the URI of the paper you link to to fetch the appropriate metadata.

If you want a standard way to say "here's the info about this post", have a look at the 'link rel="alternate"' in the head of this page.

I don't see a reason you couldn't use rev="review" more than once either, but in the postgenomic wiki you linked me to just the other day they say to use it if your post is about one paper in particular, and mention hCitation as the alternative. Perhaps I'm just misunderstanding the intention.

Your suggestion is to simply indicate that a post is or contains reviews and let the aggregator site parse things out on their end? I like the idea, because it makes things easier for content authors, however it doesn't give them any control over which bits of metadata they wish to expose.

I guess that goes back to the central issue of end user control vs. complexity.

Thanks for the link rel="alternate" example. I'm still learning this stuff.

Mr Gunn: The reason for the "one rev='review' per post" was so that an aggregator could say "_this_ piece of text is about _this_ paper". With two per post you wouldn't be able to do that so clearly - however, I think it's probably easier to do this than mandate a container around each piece of text (like hreview requires).

At the time that instruction was written I think everyone thought microformats would be a lot easier to produce than they've ended up being, and in this case it's probably easiest to avoid hreview and do the simplest thing possible.