Skip to content

lokinet v0.9.3
Compare
Choose a tag to compare
@majestrate majestrate released this 10 Jun 10:32
· 951 commits to dev since this release
v0.9.3
afe55f0

lokinet v0.9.3

v0.9.3 contains regression fixes for relays and clients, upgrade is either required or highly recommended.

relay side changes

  • properly update service node list periodically (fixes the network wide regression that makes lokinet "not work good")
  • add lokinet reachability testing and reporting to be enforced some time in the possibly near future (pending on oxend release)
  • track decommissioned service nodes as well as active service nodes, allow sessions to decommissioned service nodes but not paths, keep any existing paths open until they are dead
  • mitigate race condition on handling LinkRelayStatus Message that would not properly propagate a failure to the previous hop

client side changes

  • re-enable multithreading on clients (but not service nodes)
  • better handling of premature failures from outbound lookups so that dns fails happen only when a total fail happen. partial fails do not propagate a fail right away
  • better path handover logic
  • lower jitter from lack of a client side flush when there is low packet rate
  • path latency measurement FEC
  • path latency now uses the mean of the last N samples instead of the last measurement
  • made client side profiling super aggressive in order to prune bad/dead/stale/unreachable relays faster. will most likely result in a sharp decrease in known nodes as the dead ones are pruned
  • client side profiling decays every 30 seconds instead of every 30 minutes to speed up profiling redemption (marking possibly bad routers as good again)
  • introsets are published via near routers instead of just random ones
  • clients will now only look up outbound snapp convotags, inbound snapp convotags no longer "lookup and connect back" on failure to get an active convotag
  • reduce ONS endpoint diversity to accomidate low end devices that don't build paths as fast, this way looking up things like exit.loki happen faster
  • properly reset invalid/non-existent convotags in every failure case we can

liblokinet changes

  • fix windows build so that it (probably) produces a proper dll
  • export all functions properly on windows (probably)

misc changes

  • added new mode to lokinetmon for viewing information about introsets and flows for clients (introset inspector mode)
  • lokinetmon now uses argparse and has 2 flags: --url to set the rpc url and --introset to enter introset inspector mode
  • decrease event loop tick interval from 100ms to 250ms to reduce log spam

Hashes

[insert hashes here when they are ready]

Assets 6
Loading