After coming up with a solution to how to implement the blog:via machine tag on this site, I turned my attention this morning to how to do something interesting with it.
Since I used a custom field to implement adding machine tags to posts, I looked for a way to easily search values in those fields. I found a Wordpress plugin - Search Custom Fields - that lets me search the values in a custom field by appending key=<custom field name>.
http://redmonk.net/?key=machinetags&s=blog:via
As you can see from that uri (try it out), the search term now applies to the custom field, not the post content. But searching for all posts with ‘blog:via’ is not that interesting. If you check out this post page, I’ve tweaked the machine tags display so that next to each “blog:via” is a “more” link that takes you to a list of posts that were via that URL:
It’s not perfect yet - with my implementation (a ‘machinetags’ custom field with ‘blog:via=http://daringfireball.net’ value), I can’t actually search for “machine tag where the tag is blog:via and the value is <url>”. I may yet switch to a “blog:via” custom field.
Also, geeks ahoy! I’ve started a page on Machinetags.org for blog: machinetags.
Great work, I had a brief conversation with Jeremy at @media london about the convention for predicates and he talked about blog: then. I have followed your posts since I picked them up. With reference to the blog:xref idea is this as opposed to blog:blog, blog:postid ? If so how do you suppose handling a parser that is looking to collect all posts from a particular blogger ( I know RSS could provide this info) but my thoughts are blog:blog would be invaluable to such a parser.
My thoughts on predicates would be swayed towards blog:blog blog:postid blog:via blog:dateposted blog:author blog:url blog:permalink(not sure)
Hi Shaun,
Welcome to the conversation. In my mind, I'm trying to stay away from combinations of predicates that depend on each other for meaning. In the case of blog:blog and blog:id, I realized that (in the case of Flickr) a photo could be tagged blog:id=1234, and blog:blog=adactio.com/journal, but what if the photo is of Jeremy and you both, for example, and you want to tag it the same. how do you programatically connect the 4 tags properly then:
I simplified it to blog:xref so that the blog and post get married to a single permalink. Isn't that what myiste=myid is? Now, that photo has:
Anyway, it's a work in progress for sure.
Since I'm on a roll, I also question the long list of machine tags in specs like book... after you have that many predicates, it begine to look like the basis of a microformat instead. A microformat conceptually provides a container for all the properties, where (to me) machine tags needs to represent an encapsulated data point that can act as a sort of web "pivot-point" to link to applications.
That said, it's possible that any or all of those predicates in book are valid pivot points - I'm sort of making up this analysis as I go.
You could do the same thing for OpenID users... for filtering comments by certain users (across blogs even!):
openid:url=http://factoryjoe.com
etc.