2004-10-30 jrandom
* Cache the temporary objects used in the AES encryption/decryption
process so that AES doesn't require any memory allocation to process
data.
* Dramatically reduce memory usage within various crypto implementations
by avoiding unnecessary (though simplifying) buffers.
* If we specify some tags to be sent in an I2CP message explicitly, use
only those, not those plus a new set (otherwise we aren't sure on ACK
which set was delivered)
* Allow configuration for the partial send timeout (how long before
resending a message down a different tunnel in a lease). This can be
updated with the "router.clientPartialSendTimeout" router config prop.
* Logging
* Strip the Referer, Via, and From headers completely, rather than
inserting a bogus value ("i2p"). This should help with the use of
SnipSnap and Geeklog (thanks nickster and DrWoo!)
* Fix a strange race condition on i2cp client disconnect.
* win98 startup fixes (thanks tester-1 and ardvark!)
* include build scripts for the new streaming lib (which is NOT ready
for use yet, but you can hack around with it)
(enjoy, duck)
packets through that point have been ACKed, throwing an
InterruptedIOException if there was a writeTimeout or an IOException
if the con failed
* revamped the ack/nack field settings to ack as much as possible
* handle some strange timeout/resend errors on connection
* pass 1/2rtt as the packet 'optional delay' field, and use that to
schedule the ack time (the 'last' messages in a window set the
optional delay to 0, asking for immediate ack of all received)
* increase the optional delay to 2 bytes (#ms to delay)
* inject random failures and delays if configured to do so in
PacketHandler.choke
* fix up the window size adjustment (increment on ack, /= 2 on resend)
* use the highest RTT in the new RTT calculation so that we fit more
in (via SACK)
* fix up the SACK handling (duh)
* revise the resend time calculation
* properly close the source file in StreamSinkSend
* always adjust the rtt on ack, not just for packets with 1 send
* handle dup SYN gracefully
* revamp the default connection options
* logging
* immediately send an ack on receiving a duplicate payload message
(unless we've sent one within the last RTT)
* only adjust the RTT when there have been no resends
* added some (disabled) throttles - randomly injecting delays on
received packets, as well as randomly dropping them
* logging
has session tags within it, send an additional ping to the peer,
bundling those tags a second time, ACKing those tags on the pong.
* handle packets transferred during a race after the receiver ACKs the
connection but before the establisher receives the ACK.
* notify the messageInputStream reader on close (duh)
* new stream sink test, shoving lots and lots of data down a stream
with the existing StreamSinkServer and StreamSinkClient apps
* logging
* Allow explicit inclusion of session tags in the SDK, enabling the
resending of tags bundled with messages that would not otherwise
be ACKed.
* Don't force mode=guaranteed for end to end delivery - if mode=bestEffort
no DeliveryStatusMessage will be bundled (and as such, client apps using
it will need to do their own session tag ack/nack).
* Handle client errors when notifying them of message availability.
* New StreamSinkSend which sends a file to a destination and disconnects.
* Update the I2PSocketManagerFactory to build the specific
I2PSocketManager instance based on the "i2p.streaming.manager" property,
containing the class name of the I2PSocketManager implementation to instantiate.
* More aggressively fail peers if their tunnels are failing so that we
move off them quicker.
* Simplify some data structure serialization for reuse in the streaming
lib, as well as add support for signing and verifying partial byte
arrays.
* Logging updates
* Increased the default minimum tunnel test time to 5 seconds, since we
still see the occational message processing time spike to 2 seconds.
* Update the SimpleTimer to allow rescheduling a task thats already
queued (useful for the new streaming lib).
* Replaced old minimum tunnel test timeout of 1s with a configurable
value (router.config property "router.tunnelTestMinimum", with the
default of 2s).
* Tunnel rejection is no longer a sign of an overwhelmingly loaded
peer, so don't use it as a key point of the IsFailing calculator.
We still use it as a key point of the Capacity calculator, however.
* Allow for a configurable tunnel "growth factor", rather than trying
to achieve a steady state. This will let us grow gradually when
the router is needed more, rather than blindly accepting the request
or arbitrarily choking it at an averaged value. Configure this with
"router.tunnelGrowthFactor" in the router.config (default "1.5").
* Adjust the tunnel test timeouts dynamically - rather than the old
flat 30s (!!!) timeout, we set the timeout to 2x the average tunnel
test time (the deviation factor can be adjusted by setting
"router.tunnelTestDeviation" to "3.0" or whatever). This should help
find the 'good' tunnels.
* Added some crazy debugging to try and track down an intermittent hang.
* Fix the probabalistic tunnel reject (we always accepted everything,
since the docs on java.util.Random.nextDouble() are wrong..)
* Fixed a race on startup (thanks Quadn!)
* Disable the probabalistic drop by default (enable via the router config
property "tcp.dropProbabalistically=true")
* Disable the actual watchdog shutdown by default, but keep track of more
variables and log a lot more when it occurs (enable via the router
config property "watchdog.haltOnHang=true")
* Implement some tunnel participation smoothing by refusing requests
probabalistically as our participating tunnel count exceeds the previous
hour's, or when the 10 minute average tunnel test time exceeds the 60
minute average tunnel test time. The probabilities in both cases are
oldAverage / #current, so if you're suddenly flooded with 200 tunnels
and you had previously only participated in 50, you'll have a 25% chance
of accepting a subsequent request.
* Added a watchdog timer to do some baseline liveliness checking to help
debug some odd errors.
* Added a pair of summary stats for bandwidth usage, allowing easy export
with the other stats ("bw.sendBps" and "bw.receiveBps")
* Trimmed another memory allocation on message reception.
* Don't kill the establisher threads during a soft restart.
* Attempt to validate the peer's routerInfo earlier during handshaking.
* Revamp the AESOutputStream so it doesn't allocate any temporary objects
during its operation.
* Reimplement the I2NP reading with less temporary memory allocation.
There is still significant GC churn, especially under load, but this
should help.
* Catch some oddball errors in the transport (message timeout while
establishing).
