String representation of predicate trees.

[Leif Warner]

It looked like you'd done some work in the repo in this regard?

Some background on the use-case I was hoping to apply them to:
When the user edits a representation of the RDF "document" in their browser and hits save, I send back an HTTP PATCH (http://tools.ietf.org/html/rfc5789) of their edits to the graph in something like the talis changeset ontology.  Really nothing more than a graph of removals, and a graph of additions, to be applied atomically.  For the inverse of this, I was thinking people could call the resource with something like the HTTP header for doing a conditional GET (if-modified-since), where it returns a list of changesets applied since that date.  This could be useful for other websites wanting to stay in sync, wiki-style edit histories, or even just constructing a web view of an activity feed a la facebook, github, et al.
Then I got to thinking, rather than polling that URL, a paradigm like PubHubSubBub might be a good fit - basically, saying "post the changes back to this URL I provide when you get updates".
Like in the example registered with curl here:
http://developer.github.com/v3/repos/hooks/#pubsubhubbub
It seems to say "notify the URL at hub.callback when the subject identified by hub.topic gets updates"

What this has to do with predicate trees:
A URL like http://github.com/leifw/foo makes sense to me as an identifier for my project "foo".  But how do I specify the properties of that project I am interested in recieving upates on?  I.e. Which graph edges to I want to traverse to produce the update of what I'm interested in?
Does something like hub.topic=http://github.com/leifw/foo?paths={issues{title, description}, comments{text}} make sense for saying "I want a feed of the issues and comments of that resource", or is it silly to try to include that info in the URL?  What kind of syntax did you have in mind?  (BTW, they seem similar to SPARQL CONSTRUCT queries)
-Leif