Lots of Topics

[John Griffith <griffith@sfs.nphil.uni-tuebingen.de>]

I have a couple requests related to topic hierarchies.

1) I would like an option to only show topics if there is something
   unread-ish in it.

   Ie. 

   Topic0
     Topic1
       group1
     Topic2
     Topic3
       group2

   Do not show Topic2 here since there is nothing in it.  This could
   really cut out uninteresting screen clutter.  Of course this also
   requires a way to show all topics even if there isn't anything in
   them.

2) Many versions ago, I tried using a really large topic hierarchy.  I
   made a topic hierarchy from the first three levels of the Dewey
   Decimal Classification (between 900 and 1000 topics).  I added
   these to gnus-topic-topology and gnus-topic-alist, but I did not
   assign any groups to gnus-topic-alist (although I have worked on
   assigning the sci.* groups to DDC).  Anyway, what I noticed when I
   did that was that exiting groups and returning to the group buffer
   was extremely slow.  I have not checked the code to find out
   exactly what's going on and have not tried this in a few versions,
   but it seemed that the topic hierarchy was being rebuilt after
   exiting groups.  If this happens, is it necessary?

   If anyone wants, I can send them the DDC hierarchy.

3) What about non-unique names for topics?

   I realize this would require a new data structure to record paths
   in the topic hierarchy, but I at least would find it useful to use
   topic names more than once.  A simple example:

   Artificial Intelligence
      Programming Languages
         comp.lang.prolog
         comp.lang.lisp
   Information Systems
      Programming Languages
         comp.lang.java

   At this point the best one could do is something like:

   Artificial Intelligence
      Programming Languages - AI
         comp.lang.prolog
         comp.lang.lisp
   Information Systems
      Programming Languages - Info Sys
         comp.lang.java
  
4) What about "linked" topics?

   Ie. topic graphs rather than topic trees.  Then you could have
   symmetric topics like:

   Artificial Intelligence
      Programming Languages    [a]
   Information Systems
      Programming Languages    [b]
   Programming Languages
      Artificial Intelligence  [a]
      Information Systems      [b]

   where "Artificial Intelligence:Programming Languages" is the same
   set of groups as  "Programming Languages:Artificial Intelligence".