 Arceliar
|
22869cfffb
Use simpler 1-bit-wide buckets.
|
9 years ago |
 Caleb James DeLisle
|
cd39e99509
Improve NodeStore_brokenLink()
|
9 years ago |
|
Caleb James DeLisle
|
0af9323b69
Improve splitLinks and prevent re-pinging the same nodes over and over from nodeMill and linkMill
|
9 years ago |
|
Caleb James DeLisle
|
8724838dc9
switch to a different method of weighing additions to the external rumormill
|
9 years ago |
|
Caleb James DeLisle
|
33acce27e0
Removed last reminents of Pinned Nodes and tried to avoid adding the same nodes to the external rumormill over and over
|
9 years ago |
|
Caleb James DeLisle
|
a5117df6e3
Change Pathfinder to notify the core whenever a better path is discovered by the NodeStore, also some refactoring in the Allocator
|
9 years ago |
|
Caleb James DeLisle
|
33815d555b
tuning and improvements to --bench
|
9 years ago |
|
Arceliar
|
23cc805347
Move eventBase out of public part of nodeStore
|
10 years ago |
|
Arceliar
|
e7a8dd6b5d
Store the time each node was last pinged successfully. Reset on path changes. Include it in dumptable. Added support for this in dumptable.js
|
10 years ago |
|
Arceliar
|
b8e92c07a3
Use a 4-bit prefix per dht bucket instead of 1-bit. In a broken state at the moment.
|
10 years ago |
|
Arceliar
|
9d0fdabe17
DHT maintenance without searches. Refactored the common bit-twiddling part of the DHT stuff, so it only has to appear once.
|
10 years ago |
|
Caleb James DeLisle
|
7667736795
NodeStore_brokenLink() instead of NodeStore_brokenPath()
|
10 years ago |
|
Caleb James DeLisle
|
6f7906bc90
make sure to periodically ping existing nodes
|
10 years ago |
|
Caleb James DeLisle
|
3cfd203557
Peers are now excepted from the hard limit on number of nodes
|
10 years ago |
|
Caleb James DeLisle
|
e9aad0d066
Simplify List interface and fixed some weird behavior with the last page of dumpTable
|
10 years ago |
 Caleb James DeLisle
|
46d64b8f70
Rewrote discoverLink()
|
10 years ago |
|
Caleb James DeLisle
|
c6aef1cce8
#1 Replace NodeStore_nodeForPath() with NodeStore_linkForPath() which is more useful, #2 fix a crash bug in Janitor, #3 add a TODO comment
|
10 years ago |
|
Caleb James DeLisle
|
0f3e478936
#1 if a node 'magically changes path', flush all outgoing links from it's next-up peer, #2 squeeze a bit of logic out of NodeStore and put it other places.
|
10 years ago |
|
Caleb James DeLisle
|
14021ae0c9
Refactoring of NodeStore to use path walking instead of findClosest() with increasing hop limits.
|
10 years ago |
|
Caleb James DeLisle
|
accb8520e4
Remove the old unused capacity field from nodeStore in favor of nodeCapacity and linkCapacity
|
10 years ago |
|
Arceliar
|
76e67e6d6d
Comment cleanup
|
10 years ago |
|
Arceliar
|
d1babe5817
Interface change to NodeStore, so reach updates are done by supplying a path and latency, and all reach logic is handled in the NodeStore
|
10 years ago |
|
Arceliar
|
f76bb73ff3
Pay attention to the path a packet takes when updating reach due to a ping/search response.
|
10 years ago |
|
Caleb James DeLisle
|
f70518c9d3
bound the size of the nodestore
|
10 years ago |
|
Caleb James DeLisle
|
1f1cf83d4e
Work from over the weekend, crashey is less crashey (at least in the sim)
|
11 years ago |
|
Caleb James DeLisle
|
e28592a331
NodeStore_discoverNode() needs to continue splitting links based on the result of the first split.
|
11 years ago |
|
Caleb James DeLisle
|
094e5b94e0
Fixed a few more bugs
|
11 years ago |
|
Caleb James DeLisle
|
2c25a476b3
Moved the code over to completely use the new pathfinder
|
11 years ago |
|
Caleb James DeLisle
|
4c794eb053
A little bit less crashey
|
11 years ago |
|
Caleb James DeLisle
|
4ed88710b5
way more work on the new pathfinder
|
11 years ago |
