===================================================================
RCS file: /cvs/cvsweb/TODO,v
retrieving revision 3.1.2.1
retrieving revision 4.3
diff -u -p -r3.1.2.1 -r4.3
--- cvsweb/TODO	2019/11/07 09:44:35	3.1.2.1
+++ cvsweb/TODO	2019/11/11 13:15:09	4.3
@@ -1,44 +1,38 @@
-$Id: TODO,v 3.1.2.1 2019/11/07 09:44:35 schwarze Exp $
-$knu: TODO,v 1.1 2000/07/19 21:59:47 knu
+$Id: TODO,v 4.3 2019/11/11 13:15:09 schwarze Exp $
+$knu: TODO,v 1.10 2004/10/31 22:46:00 scop
 
-o request tar or zip of entire directory (any tag); caching
-o limit the number of logentries shown
-o show left/right/up icons in diff/logtable
-o implement frames (backbutton on top of page)
-o DiffSelection: jump back to 'User Text Field'
-  if entered text (javascript - anyone a javascript
-  programmer ?) - 
-  done, thanks to Andy Sharp <asharp@hobbes.clarity.ca>
-o make 'exec's save: use full path and do not spawn shell
-o use strict, Apache: mod_perl; done, thanks to
-  Jens-Uwe Mager <jum@helios.de>
-o explore if we can use cvs directly to abandon the
-  use of 'rlog', 'rcsdiff'. This could be done much
-  like it is for cvs annotate; 
-  'co' done, but it seems to have problems with files which
-  have spaces in the path (rare) -> CVS bug
-  The 'rlog'-function could be implemented just like the 
-  annotate function, using a cvs in server mode. 
-  The 'rcsdiff'-functionality is almost covered by 
-  'cvs rdiff' .. but I am missing the -p and -w options
-  then ..
-  Any way to traverse the directory structure with cvs C/S 
-  ( we need the cvs list patch here I think) ? 
-  The performance could be - C/S-mode - .. bad. Caching ?
-  This way we someday won't need to have physical access to
-  the Repository (though this may not be what we want: consider
-  10'000 Users browsing your anonymous CVS via a remote cvsweb)
-o try to figure out what's the latest change within a subdirectory ..
-  the trivial (recursive ..) solution would be a bit timeconsuming
-o include date in the checkout pathinfo (+branchname),
-  to get really what you want (the current solution is not optimal
-  if you manage/review webpages: you always get a checkout of the
-  _latest_ pictures embedded in the webpage, not pictures from the 
-  same date the page is)
-o rewrite this stuff as servlet ..
-o check wether we can set $HOME to /tmp - to avoid problems with
-  cvs looking for .cvsrc etc.
-o annotate with last Log-Message in the header
-o allow to choose which representation we have in dir-view
-o open external window only if we've a specific mime-type (html)
+TODO list for CVSweb
+-------------------------------------------------------------------------------
 
+o Tear the code apart, and modularize.
+
+o Ensure that fatal error messages are shown properly when gzip encoding
+  is enabled.
+
+o Use specialized modules from CPAN to do stuff, maybe these would be useful:
+  + Archive::Tar
+  + Archive::Zip
+  + Compress::Zlib
+  + Compress::Bzip2
+  + HTML::Entities
+  + Cvs or VCS::LibCVS
+
+o Integrate/cooperate with various CVS utilities.
+  + cvs2cl      <http://www.red-bean.com/cvs2cl/>
+  + cvsplot     <http://cvsplot.sourceforge.net/>
+  + CVSps       <http://www.cobite.com/cvsps/>
+
+o Limit the number of log entries shown.
+
+o Explore whether we can use cvs for all operations and drop rlog, rcsdiff
+  and friends.
+
+o Find out if it is feasible to provide prune-like functionality, ie. hide
+  empty subdirs.
+
+o Recursive diffs, including support for tags/branches/dates.
+
+o Support for dates in tarballs/zips (in addition to tags/branches).
+
+o Make it possible to pass media type parameters with the content-type
+  query string parameter, for example "Content-Type: text/html; charset=utf-8".