Motivation
So, why do this thing?
This is meant as another piece in reducing platform partisanship. Right now you have to choose your platform, then seek software written for that platform. That platform can include your operating system, database, front-end web server. But mostly it is your development language, web framework, templating language, and authentication system.
WSGI helps mitigate some of the framework differences, if not the language differences (though with things like paste.proxy it can be moderately language-neutral). This doesn't remove the problems of templating languages. Many frameworks (or at least applications) include ways of overriding templates in site-specific ways. But each of these is typically bound to the framework, or at least the templating language. Basic composition of a web site from multiple applications is difficult, and stylistic choices have to be repeated across all your different platforms. CSS helps a little, but just a little.
One Templating Language To Rule Them All seems unrealistic, especially because it is so closely tied to development language (things like Clearsilver aside). Also, these typically deal with objects bound to their respective implementation languages (or require sometimes-painful mapping as with Clearsilver).
There's already a solution to this, of course: XSLT. Everyone loves XSLT! Well, that's not true, but this attempts to do the same thing as XSLT, but not as general, and very much focused on HTML.
There's some other things that are less markup-focused: server-side includes and edge-side includes. SSIs have some of the same problems that other templating languages have; even if the data model is primitive, it is still there. Nevertheless, with Apache 2 and output filtering, SSIs are capable. ESIs are more network oriented, but pretty much dead. These are both textual languages; they don't attempt to understand the structure of the page at all.
This project is another run at the same idea. This time the project is only focused on HTML. It has special handling for tags in <head>, for instance, to assemble a page with all the combined prerequesites (scripts and stylesheets) of its component pieces
While currently this is build on Python and WSGI, nothing in this is actually bound to these pieces, they are just a good platform for experimenting with these ideas. The templating itself is language neutral, and could be implemented in any language.
So How Can This Really Succede?
After all that, isn't this just another templating language?
Yes and no. Yes, in that there are no applications out there that use this. Unlike XSLT, this is not so powerful that you could apply this to an unwilling application (though I doube you can realistically do that with XSLT either).
The idea is to use the template overriding facilities that applications and frameworks already have, to apply an overlay template. This overlay template could be used by anyone to apply their styles; so even though it requires per-application work, that work can be done by anyone and shared with everyone.