Theo (and others), I've been thinking about the URI code.  Right now,
the keys are the URIs as parsed.  I have not really liked that as it's
not really a straightforward representation of the message contents,
etc.  I proposed using an array of hashes where each array element is
one URI.

The problem with even that proposal is this:

 - we parse URIs during HTML rendering
 - we then grab URIs from the message text

However, we don't do much to associate the two with each other, etc.
Ideally, I think we would have that array, but it would cover both text
and HTML -- if the part was HTML, all the URI stuff (both non-hyperlink
text of URIs and in HTML URIs that are hyperlinks) would be done at HTML
render time and if it's text, it would be done during text rendering.

Another idea I had is this: use a hash like we do currently, but have
the key be the canonical URI only, not the original URI, to avoid
replication, etc.  Then the value hash would be various things like
where it came from, the original forms, etc.

What do you guys think?

Daniel

-- 
Daniel Quinlan
http://www.pathname.com/~quinlan/

Reply via email to