I had a look at dojo before and these were the questions that popped up in my mind:
Dojo seems to rely on adding its own unnamespaced attributes to normal elements. Is this compatible with strict use of XHTML? What would the mechanism for fallbacks be? To produce the non-JS case, then use unobtrusive JS to add dojos funny attributes, and then trigger dojo to turn it into a dojo widget, or what? What 'profile' of Dojo would we bundle? How would it be handled in our svn? A script to go and fetch it from Dojos svn and run the compressor, and something to complain if this hasn't been done? Use this to create the scripts to include in tarballs? I don't think we want to mirror Dojo in our svn, but rather just track a tagged release? I think there are a couple of other things that could be interesting here... I think #925 is very important. A 'component' effectively becomes a template-tag + a processor. This can be used to create a 'component' generic view for partial page updates. Needs more thought (preserving other component state on normal posts), but could be amazing if we can pull it off. on my local parser-state branch (see the message Parser Pranks) , I'm now experimenting with dynamically creating an ElementTree from templates as I go along. This can trivially be transformed to HTML4 or Mochi-esque JSON DOM on output. This might give us some flexibility on the transport side.
