Hi Cedric:

Thanks for confirming my thoughts on how the matching code works.

We support multiple dialects of markdown the same way we support
multiple templating languages. Different dialects have different
features. In theory it is possible to override the import function in 
interfaces.py so a site could use a custom configuration of one
of the markdown back ends with the features/plugins they need.

Yup testing the common subset is a pain because the language isn't 
specified in a way that maps input to output. Ideally every markdown 
implementation would produce the same html. But ...

Testing the output from the markdown process allows us to verify
that the processing pipeline is producing expected html for
presentation to the user. But it is a pain.

Regarding email, have you looked at what markdown2 emits? Mistune
puts out:

  <a href="mailto:user@example.com">user@example.com</a>

while markdown2 seems to have an ever changing set of entities:

  href="mai ..."

and sometimes they throw in actual characters.

markdown uses decimal entities and is the same from run to run.

So yeah you have convinced me that validating the email matching
path is more trouble than it's worth.