You can clone with
HTTPS or Subversion.
What is needed? First stab at the requirements is: Motivation and Requirements
Please add any thoughts, comments, etc.
Do you know the Norman Walsh resolver project? https://github.com/ndw/xmlresolver
It seems to have similar goals.
Unfortunately Java SDK does not lack resolver interfaces. org.w3c.dom.ls.LSResourceResolver is another one.
Why not write a resolver that enables resolution to resources on the classpath? I mean why does everyone assume that entities need to be first downloaded and then cached, or else located in an OS dependent location. Most applications have predictable requirements i.e. only require HTML, XHTML or some other fixed set of DTDs. These could easily be packaged into JARs if such a resolver also exists that can read them.
Just a thought, perhaps naive somehow but you asked for thoughts ;-)
I appreciate that (was away, still catching up on mails). Yes, classpath is an interesting option. I will consider it further.
OK, welcome back :-)