solid nnweb

[Lars Magne Ingebrigtsen <larsi@ifi.uio.no>]

The ephemeral nnweb groups are nice and fun, but I think solid nnweb
groups would be actually useful.

Imagine -- you create an nnweb group that matches "Mango".  nnweb
would keep an identifier<->article number mapping somewhere, so that
when you tap `M-g' on the nnweb:mango group, it would request matches
from the search engine, compare with its previously created map, and
make all the new articles, well, new.  In short, solid nnweb groups
could potentially behave like any other group.

I see two possibilities for implementing this.

1) Just keep a simple identifier-to-article number map in a file.  The
identifier would be either the (partial) URL to fetch the article or
the Message-ID.  (Depends on what data the search engine gives.)  This
would simply make read articles read and unread articles unread. 

2) Keep all the headers to all the previous search results stored in a
file.  Then commands like `C-u RET' will let you enter a summary
buffer without actually consulting the search engine.  Selecting an
article would mean fetching from the search engine site, but that's a
different matter.

Hmm...  The second alternative seems sexier.  How would old entries be
expired?  Just keep a max-size variable, and expire the oldest
entries?  I suppose that could be done...  

Ok: There would be a ~/News/nnweb/ directory with an active file, a
file to store the match regexp (and stuff) for the groups, and one
.overview file per nnweb group.  Uhm -- the active file and the regexp
file could be combined, I guess.  Yes.

So -- whaddaya think?  Ephemeral nnweb groups would work as present,
of course. 

-- 
  "Yes.  The journey through the human heart 
     would have to wait until some other time."