1 PHASE #1: (Goal: settle key design questions)
4 * Windows: use events instead of pipes to signal select()s [Nils]
5 * only connect() sockets that are ready (select()) [Nils]
6 [On W32, we need to select after calling socket before
9 PHASE #2: (Goal: recover basic file-sharing functionality)
14 For these functions, it would be nice if we had testcases ("make check")
15 that would cause them to be executed and check that they are working:
16 * gnunet-service-peerinfo:
17 - change_host_trust / flush_trust
19 - discard_hosts_helper / cron_clean_data_hosts
20 * gnunet-service-transport:
21 - try_unvalidated_addresses
22 - lookup_address_callback
23 - lookup_hello_callback
24 - plugin_env_lookup_address
25 - notify_clients_disconnect
26 - list_validated_addresses
28 - disconnect_neighbour
30 * plugin_transport_tcp.c:
32 - tcp_plugin_address_pretty_printer / append_port
33 - tcp_plugin_set_receive_quota
36 - GNUNET_TRANSPORT_set_qutoa / send_set_quota
40 - remove_from_any_list / remove_neighbour
41 - GNUNET_TRANSPORT_notify_transmit_ready_cancel
42 - Testcases for set_quota, timeouts, disconnects, transmit_ready_cancel
43 - gnunet-service-transport HELLO validation (how good is our coverage?)
44 - direct test of plugins compliance to plugin API
47 - solicit_traffic / copy_and_free
48 - GNUNET_CORE_peer_configure / produce_configure_message
49 * gnunet-service-core:
52 - handle_client_request_configure
55 - handle_transport_notify_disconnect
56 * hostlist (everything)
57 * topology (everything)
60 Module features to implement:
61 * Dstore (needed for DHT)
63 - design network structs
65 - implement sqlite-based dstore plugin
66 - implement dstore service
67 - implement dstore API
68 - implement testcases for dstore API
69 - implement testcases for dstore plugin
70 - implement performance tests
71 * TESTING (needed for DV, DHT, Topology)
73 - implement library for local testing
74 - implement testcases for library
75 - implement library for distributed testing
76 - implement testcases for distributed testing
78 - implement testcases (needs TESTING)
83 - implement DV service
84 - implement DV library
85 - implement DV transport plugin
87 - implement performance tests
90 - implement DHT service (needs DV, DSTORE)
91 - implement DHT library
93 - implement performance tests
96 - design network structs (CS)
97 - implement FS library
100 + persistence mechanism
110 - design network structs (P2P)
111 - datastore request queueing mechanism
112 - implement FS service (needs DHT)
117 - implement testcases
120 + sharing API (needs TESTING)
129 - implement performance tests (needs TESTING)
137 PHASE #3: (Goal: ready for pre-release) [completion-goal: end of 2009]
139 Module features to implement:
142 - need to settle basic design; do we want to keep guile?
144 - good to have for DV evaluation!
146 - good to have for DV/DHT evaluation!
156 Plugins to implement:
157 * UDP backend (need LIBRARY to support (de)fragmentation!)
164 Minor TODO items / known bugs:
166 - crypto_hash: use libgcrypt (supports SHA-512 since 2003)
167 - container_bloomfilter: improve efficiency (see FIXME)
169 - inefficient memmove
171 - transport_api: support forcing disconnects through low quotas!
172 (required for working F2F support!)
173 - API: consider having core provide deadline information for each message
174 (likely important for DV plugin which wants to loop back!)
175 - implement transport API to pretty-print transport address
176 + transport_api extension (API extension!)
177 + service-transport extension (protocol extension)
178 - add calls to statistics in various places
179 - implement gnunet-transport (transport configurator / tester)
180 - UPnP-based IP detection
181 (Note: build library always, build service when libxml2/etc. are available)
182 - instantly filter addresses from *other* peers that
183 are *equal* to our own address + port (i.e., localhost:2086). We
184 no longer filter those for outgoing (helps with loopback testing
185 and keeps the code clean), but we should filter strictly *impossible*
186 incoming addresses! This is for efficiency, not correctness.
187 - We currently are happy to take any address told to us in a WELCOME
188 to our set of addresses; we should have some minimal threshold-based
189 scheme, limiting both the total number of addresses that we accept
190 this way as well as requiring multiple confirmations; also, we
191 should possibly try to confirm that the given address works for
192 us ourselves (loopback-style) before adding it to the list
194 + we may be able to simplify WELCOME messages (no need to add
195 addresses there anymore, but may help to learn them there anyway...).
196 + we probably want some kind of voting/counting for learning IP addresses
197 (maybe including IP addresses in ads proportional to how often others
198 report them? we at least need some protection against >64k HELLOs!),
199 + provide a way to give the user a list of "learned" IP addresses and
200 a way to easily "veto" addresses off the list!
201 => If MiM attacker uses vetoed address, blacklist the specific IP for
202 the presumed neighbour!
203 - not sure current way of doing ACKs works well-enough
204 with unreliable transports where the ACK maybe lost;
205 the "is_new" check would then possibly prevent future
206 ACKs to be delivered, all while we're happily
207 receiving messages from that peer! Worse, the other
208 peer won't generate another ACK since it thinks we're
209 connected just fine...
211 + How necessary is ACKing in the first place? (alternatives?)
212 + Should we transmit ACKs in response to every HELLO? (would that
213 fully address the problem?)
214 - latency measurements implemented in the transport
215 plugins makes it only work for bi-di transports
216 and results in code replication
217 - should latency be included in the ReceiveCallback and
218 NotifyConnect or passed on request?
219 - FIXME's with latency being simply set to 0 in a few places
220 - Memory leak (running valgrind --trace-children=yes on test_transport_api:
221 ==28393== 16 bytes in 1 blocks are indirectly lost in loss record 1 of 5
222 ==28393== at 0x4C2260E: malloc (vg_replace_malloc.c:207)
223 ==28393== by 0x52343E3: GNUNET_xmalloc_unchecked_ (common_allocation.c:62)
224 ==28393== by 0x5234389: GNUNET_xmalloc_ (common_allocation.c:53)
225 ==28393== by 0x524458A: GNUNET_NETWORK_socket_create_from_accept (network.c:289)
226 ==28393== by 0x524B2DA: ??? (server.c:332)
227 ==28393== by 0x524A4C7: ??? (scheduler.c:425)
228 ==28393== by 0x524A73D: GNUNET_SCHEDULER_run (scheduler.c:510)
229 ==28393== by 0x524FF8C: GNUNET_SERVICE_run (service.c:1326)
230 ==28393== by 0x405500: main (gnunet-service-transport.c:2645)
232 ==28393== 65,744 (65,728 direct, 16 indirect) bytes in 1 blocks are definitely lost in loss record 5 of 5
233 ==28393== at 0x4C2260E: malloc (vg_replace_malloc.c:207)
234 ==28393== by 0x52343E3: GNUNET_xmalloc_unchecked_ (common_allocation.c:62)
235 ==28393== by 0x5234389: GNUNET_xmalloc_ (common_allocation.c:53)
236 ==28393== by 0x524473E: GNUNET_NETWORK_socket_create_from_accept (network.c:323)
237 (rest of trace identical)
243 - auto-generate "defaults.conf" using gnunet-setup from "config.scm"
244 - integrate all options into "config.scm"
245 - change config-file writing to exclude options set to default values
247 - implement exponential back-off for service restarts
248 - better tracking of which config changes actually need to cause process restarts by ARM.
249 - have way to specify dependencies between services (to manage ARM restarts better)
250 - client-API is inefficient since it opens a TCP connection per service that is started
251 (instead of re-using connections).
253 - code currently notifies clients about "encrypted" connections being up well before
254 we get the encrypted PONG; sometimes this may be OK (for topology killing
255 unwanted connnections), but of course not in general. I suspect we want
256 to signal on PONG and have topology hook directly into transport to
257 kill plaintext connections before they have a chance to become encrypted
258 (may require minor hack in transport API)
260 - have gnunet-peerinfo print actual host addresses again
261 - add option to gnunet-peerinfo to modify trust value
263 - finish postgres implementation; simplify other SQLs using new stats
268 - active reply route caching design & implementation of service,
271 - implement advertising of hostlist URL
272 - implement learning of hostlist URLs
278 PHASE #4: [completion-goal: mid 2010]
283 - update webpage documentation
285 - expand bibliography
286 - convert documentation pages to books
287 - update books (especially for developers)
288 - add content type for links and view
289 - create good Drupal theme for GNUnet
290 - make a NICE download page and figure out how to
291 enable developers to publish TGZs nicely
292 - port "contact" page; add impressum
293 - add content type for "todo" items?
298 * Features eliminated from util:
299 - threading (goal: good riddance!)
300 - complex logging features [ectx-passing, target-kinds] (goal: good riddance!)
301 - complex configuration features [defaults, notifications] (goal: good riddance!)
302 - network traffic monitors (goal: eliminate)
303 - IPC semaphores (goal: d-bus? / eliminate?)
305 - DNS lookup (goal: have async service; issue: still need synchronous resolution in places, current code may not be portable)
306 => code shrunk from 61 files to 34, 22k LOC to 15k LOC, 470k to 330k (with symbols)
307 * New features in util:
309 - service and program boot-strap code
310 * Major changes in util:
311 - more expressive server (replaces selector)
313 - how to integrate scheduler with GTK event loop!