flush
[oweals/gnunet.git] / TODO
diff --git a/TODO b/TODO
index ab1d9617237b597414b7928f901ca3be532c5fc1..ad814b9473ed13fe40b36f444f3d7a80c0f820b5 100644 (file)
--- a/TODO
+++ b/TODO
@@ -8,46 +8,22 @@ away), in order in which they will likely be done:
 * TRACEKIT
 * FRAGMENTATION [Ji Lu]
 * HTTP transport
-* MySQL / Postgres plugins (datastore, datacache) [Radhika]
+* MySQL / Postgres plugins (datastore, datacache)
 * UPNP [Milan]
+* ARM [Safey]
 
 Urgent items (before announcing ng.gnunet.org):
-* topology
-  - valgrind crash (in topology test case):
-==31107== Invalid read of size 8                                                                                                             
-==31107==    at 0x401DA1: free_peer (gnunet-daemon-topology.c:480)                                                                           
-==31107==    by 0x4030A2: cleaning_task (gnunet-daemon-topology.c:1212)                                                                      
-==31107==    by 0x545DCBD: run_ready (scheduler.c:482)                                                                                       
-==31107==    by 0x545E0CD: GNUNET_SCHEDULER_run (scheduler.c:609)                                                                            
-==31107==    by 0x545B00F: GNUNET_PROGRAM_run (program.c:223)                                                                                
-==31107==    by 0x403477: main (gnunet-daemon-topology.c:1346)                                                                               
-==31107==  Address 0x737e828 is 8 bytes inside a block of size 144 free'd                                                                    
-==31107==    at 0x4C21DBC: free (vg_replace_malloc.c:325)                                                                                    
-==31107==    by 0x544240B: GNUNET_xfree_ (common_allocation.c:144)                                                                           
-==31107==    by 0x401E13: free_peer (gnunet-daemon-topology.c:487)                                                                           
-==31107==    by 0x4030A2: cleaning_task (gnunet-daemon-topology.c:1212)                                                                      
-==31107==    by 0x545DCBD: run_ready (scheduler.c:482)                                                                                       
-==31107==    by 0x545E0CD: GNUNET_SCHEDULER_run (scheduler.c:609)                                                                            
-==31107==    by 0x545B00F: GNUNET_PROGRAM_run (program.c:223)                                                                                
-==31107==    by 0x403477: main (gnunet-daemon-topology.c:1346)                                                                               
-==31107==                                                                                                                                    
-  - considers peers 'connected' well before they actually are
-    (since core notifies about it too early?);
-  - (forced) disconnect does not work (also CORE API issue)
-  - needs testing (not sure the current testcase does much...)
-* hostlist
-  - test fails (looks like it works, but that's because of a bad
-    connectivity notification; somehow core is unable to send
-    messages successfully via transport)
+* TRANSPORT:
+  - nondeterministic transport testcase failures [Nate]
 * CORE:
-  - core notifies about connects "too early" (when we have not yet
-    succeeded with the full key exchange) [see also: BUGS]
+  - test currently fails
+  - request connect not working [Christian, need transport first]
+* TOPOLOGY:
+  - needs testing [need transport first]
+* HOSTLIST:
+  - needs testing [need transport first]
 * FS (basic anonymous FS only)
-  - implement FS service (P2P operations)
-    + how to send queries (soliciting is not there in core; do we
-      also want to do pushing sometimes?)
-    + need to bound queueing of replies for other peers    
-  - test multi-peer search/download
+  - test multi-peer search/download [needs core first]
 * new webpage
   - run peer => have a 0.9.x hostlist
   - improve basic documentation (configure, dependencies, what works, etc.)
@@ -61,12 +37,14 @@ Urgent items (before announcing ng.gnunet.org):
   - trust: need *fast* way to check/update trust in peers
            (async peerinfo would not be right)
   - OS: existing waitpid call is not nice (not integratable with scheduler! fix this!)
+    => should 'scheduler' support SIGCHILD explicitly? Always?
 * STATISTICS:
-  - synchronous/asynchronous API (& implementation) is not nice
+  - synchronous/asynchronous API (& implementation) is not nice;
+    => provide notification-based API
   - does not seem to work with timeouts (especially if service is not running)
 * ARM:
   - need to get rid of synchronous API for service starts (cause all kinds of problems)
-    [=> eliminate for need to tell ARM about service starts most of the time!]
+    [=> eliminate for need to tell ARM about service starts most of the time!] [Safey]
 * HELLO:
   - need function to test "equivalency" of HELLOs; use in topology!
 * Module features to implement:
@@ -85,8 +63,8 @@ Urgent items (before announcing ng.gnunet.org):
     + gnunet-download (many options)
     + gnunet-directory (man page, options)
     + gnunet-pseudonym (all of it)
-    + gnunet-service-fs (remove failing on-demand blocks, many other nitpicks/features/optimizations)
-    + datastore: do active migration support here?
+    + gnunet-service-fs (remove failing on-demand blocks, stats, hot-path routing, load-based routing, nitpicks)
+    + datastore: do active migration support here?   
   - implement adv. FS testcases 
     + getopt API
     + insert: sblocks, loc uris
@@ -190,3 +168,8 @@ Urgent items (before announcing ng.gnunet.org):
   (add options to gnunet-search / gnunet-download to limit search to local peer)
 * STATISTICS:
   - should use BIO instead of mmap
+* PEERINFO:
+  - expire 'ancient' HELLOs (those without valid addresses AND that 
+    we have not 'used' (for their public keys) in a while; need a way
+    to track actual 'use')
+  - make sue we also trigger notifications whenever HELLOs expire