* Tweak the PRNG logging so it only displays error messages if there are
problems
* Disable dynamic router keys for the time being, as they don't offer
meaningful security, may hurt the router, and makes it harder to
determine the network health. The code to restart on SSU IP change is
still enabled however.
* Disable tunnel load testing, leaning back on the tiered selection for
the time being.
* Spattering of bugfixes
* i2psnark: Implement basic partial-piece saves across connections
* i2psnark: Implement keep-alive sending. This will keep non-i2psnark clients
from dropping us for inactivity but also renders the 2-minute transmit-inactivity
code in i2psnark ineffective. Will have to research why there is transmit but
not receive inactivity code. With the current connection limit of 24 peers
we aren't in any danger of keeping out new peers by keeping inactive ones.
* i2psnark: Increase CHECK_PERIOD from 20 to 40 since nothing happens in 20 seconds
* i2psnark: Fix dropped chunk handling
* i2psnark: Web rate report cleanup
* i2psnark: Report cleared trackerErr immediately
* i2psnark: Add trackerErr reporting after previous success; retry more quickly
* i2psnark: Set up new connections more quickly
* i2psnark: Don't delay tracker fetch when setting up lots of connections
* i2psnark: Reduce MAX_UPLOADERS from 12 to 4
* Limit form size in SusiDNS to avoid exceeding a POST size limit on postback
* Print messages about addressbook size to give better overview
* Enable delete function in published addressbook
* Actually fix the threading deadlock issue in the netDb (removing
the synchronized access to individual kbuckets while validating
individual entries) (thanks cervantes, postman, frosk, et al!)
* Cut down NTCP connection establishments once we know the peer is skewed
(rather than wait for full establishment before verifying)
* Removed a lock on the stats framework when accessing rates, which
shouldn't be a problem, assuming rates are created (pretty much) all at
once and merely updated during the lifetime of the jvm.
* When dropping a netDb router reference, only accept newer
references as part of the update check
* If we have been up for a while, don't accept really old
router references (published 2 or more days ago)
* Drop router references once they are no longer valid, even if
they were allowed in due to the lax restrictions on startup
* Every time we create a new router identity, add an entry to the
new "identlog.txt" text file in the I2P install directory. For
debugging purposes, publish the count of how many identities the
router has cycled through, though not the identities itself.
* Cleaned up the way the multitransport shitlisting worked, and
added per-transport shitlists
* When dropping a router reference locally, first fire a netDb
lookup for the entry
* Take the peer selection filters into account when organizing the
profiles (thanks Complication!)
* Avoid some obvious configuration errors for the NTCP transport
(invalid ports, "null" ip, etc)
* Deal with some small NTCP bugs found in the wild (unresolveable
hosts, strange network discons, etc)
* Send our netDb info to peers we have direct NTCP connections to
after each 6-12 hours of connection uptime
* Clean up the NTCP reading and writing queue logic to avoid some
potential delays
* Allow people to specify the IP that the SSU transport binds on
locally, via the advanced config "i2np.udp.bindInterface=1.2.3.4"
2006-07-18 jrandom
* Add a failsafe to the NTCP transport to make sure we keep
pumping writes when we should.
* Properly reallow 16-32KBps routers in the default config
(thanks Complication!)
* Improve the multitransport shitlisting (thanks Complication!)
* Allow routers with a capacity of 16-32KBps to be used in tunnels under
the default configuration (thanks for the stats Complication!)
* Properly allow older router references to load on startup
(thanks bar, Complication, et al!)
* Add a new "i2p.alwaysAllowReseed" advanced config property, though
hopefully today's changes should make this unnecessary (thanks void!)
* Improved NTCP buffering
* Close NTCP connections if we are too backlogged when writing to them
* New NIO-based tcp transport (NTCP), enabled by default for outbound
connections only. Those who configure their NAT/firewall to allow
inbound connections and specify the external host and port
(dyndns/etc is ok) on /config.jsp can receive inbound connections.
SSU is still enabled for use by default for all users as a fallback.
* Substantial bugfix to the tunnel gateway processing to transfer
messages sequentially instead of interleaved
* Renamed GNU/crypto classes to avoid name clashes with kaffe and other
GNU/Classpath based JVMs
* Adjust the Fortuna PRNG's pooling system to reduce contention on
refill with a background thread to refill the output buffer
* Add per-transport support for the shitlist
* Add a new async pumped tunnel gateway to reduce tunnel dispatcher
contention
* Ensure that the I2PTunnel web interface won't update tunnel settings
for shared clients when a non-shared client is modified
(thanks for spotting, BarkerJr!)
* Cut down the proactive rejections due to queue size - if we are
at the point of having decrypted the request off the queue, might
as well let it through, rather than waste that decryption
* Only send netDb searches to the floodfill peers for the time being
* Add some proof of concept filters for tunnel participation. By default,
it will skip peers with an advertised bandwith of less than 32KBps or
an advertised uptime of less than 2 hours. If this is sufficient, a
safer implementation of these filters will be implemented.
* Fix some oversights in my previous changes:
adjust some loglevels, make a few statements less wasteful,
make one comparison less confusing and more likely to log unexpected values
* Separate growth factors for tunnel count and tunnel test time
* Reduce growth factors, so probabalistic throttle would activate
* Square probAccept values to decelerate stronger when far from average
* Create a bandwidth stat with approximately 15-second half life
* Make allowTunnel() check the 1-second bandwidth for overload
before doing allowance calculations using 15-second bandwidth
* Tweak the overload detector in BuildExecutor to be more sensitive
for rising edges, add ability to initiate tunnel drops
* Add a function to seek and drop the highest-rate participating tunnel,
keeping a fixed+random grace period between such drops.
It doesn't seem very effective, so disabled by default
("router.dropTunnelsOnOverload=true" to enable)
* Allow a single build attempt to proceed despite 1-minute overload
only if the 1-second rate shows enough spare bandwidth
(e.g. overload has already eased)
* Correct a misnamed property in SummaryHelper.java
to avoid confusion
* Make the maximum allowance of our own concurrent
tunnel builds slightly adaptive: one concurrent build per 6 KB/s
within the fixed range 2..10
* While overloaded, try to avoid completely choking our own build attempts,
instead prefer limiting them to 1
* Adjust the tunnel build timeouts to cut down on expirations, and
increased the SSU connection establishment retransmission rate to
something less glacial.
* For the first 5 minutes of uptime, be less aggressive with tunnel
exploration, opting for more reliable peers to start with.
