99 lines
3.6 KiB
HTML
99 lines
3.6 KiB
HTML
<pre>-----BEGIN PGP SIGNED MESSAGE-----
|
|
Hash: SHA1
|
|
|
|
'lo all, time for the weekly rant session
|
|
|
|
* Index:
|
|
1) 0.3.3 & current updates
|
|
2) NativeBigInteger
|
|
3) ???
|
|
|
|
* 1) 0.3.3
|
|
|
|
We pushed out the 0.3.3 release this past Friday and after a day
|
|
or two of pretty bumpy weather, it seems to be doing ok. Not as
|
|
good as 0.3.2.3, but I've usually been able to hang around on
|
|
irc.duck.i2p for 2-7h stints. However, as I've seen lots of
|
|
people having trouble, I fired up the logger and monitored in
|
|
detail what was up. Short answer is that we were simply using
|
|
more bandwidth than we need to, causing congestion and tunnel
|
|
failures (due to test messages timing out, etc).
|
|
|
|
I've spent the last few days back in the simulator, running a
|
|
series of heartbeats through a network to see what we can
|
|
improve, and we've got a whole slew of updates coming our way
|
|
based on that:
|
|
|
|
= netDb update to operate more efficiently
|
|
The existing netDb lookup messages are up to 10+KB, and while
|
|
successful replies are frequent, the unsuccessful replies
|
|
could be up to 30+KB (as both contained full RouterInfo
|
|
structures). The new netDb replaces those full RouterInfo
|
|
structures with the router's hash - turning 10KB and 30KB
|
|
messages into ~100 byte messages.
|
|
|
|
= throw out the SourceRouteBlock and SourceRouteReplyMessage
|
|
These structures were a remainder of an old idea but don't add
|
|
any value to the anonymity or security of the system. By
|
|
dropping them in favor of a simpler set of reply data points,
|
|
we cut the tunnel management message sizes dramatically, and
|
|
drop the garlic encryption time in half.
|
|
|
|
= removed some excess messages
|
|
The code was a bit 'chatty' during the tunnel creation, so the
|
|
unnecessary messages have been cut.
|
|
|
|
= reduced arbitrary padding
|
|
Some of the crypto code for the garlic routing was using fixed
|
|
padding based on some garlic routing techniques that we're not
|
|
using (when I wrote it back in September and October I thought
|
|
we were going to be doing multi-hop garlic routing instead of
|
|
tunnels).
|
|
|
|
I'm also working on seeing if I can get the full blown update
|
|
to the tunnel routing to add the per-hop tunnel ids.
|
|
|
|
As you can see from the roadmap, this encompases a lot of the
|
|
0.4.1 release, but since the netDb change meant losing
|
|
backwards compatability, we might as well get a slew of
|
|
backwards incompatible things done at once.
|
|
|
|
I'm still running tests in the sim and have to see if I can
|
|
finish up the per-hop tunnel id thing, but I hope to have a
|
|
new patch release out in a day or two. It won't be backwards
|
|
compatible, so it'll be a bumpy upgrade, but it should be
|
|
worth it.
|
|
|
|
* 2) NativeBigInteger
|
|
|
|
Iakin has been doing some updates to the NativeBigInteger code
|
|
for the Freenet team, optimizing some stuff we don't use, but
|
|
also putting together some CPU detection code that we can use
|
|
to automatically select the right native library. That means
|
|
we'll be able to deploy jbigi in a single lib with the default
|
|
install and it'll pick the right one without having to ask the
|
|
user for anything. He has also agreed to release his mods and
|
|
the new CPU detection code so that we can bundle it into our
|
|
source (yay Iakin!) I'm not sure when this will be deployed,
|
|
but I'll let people know when it is, as those with existing
|
|
jbigi libraries will likely need a new one.
|
|
|
|
* 3) ???
|
|
|
|
Well, the last week has been a lot of head in the code hacking,
|
|
so not too many updates. Anyone have anything else to bring
|
|
up? If so, swing on by the meeting tonight, 9p GMT in #i2p.
|
|
|
|
=jr
|
|
|
|
-----BEGIN PGP SIGNATURE-----
|
|
Version: PGP 8.1
|
|
|
|
iQA/AwUBQQavYhpxS9rYd+OGEQKiewCg+Bcfv1tTkvMkk6yRnfwKnAKniEgAoN7z
|
|
Q36Vr3muI4ti770dlw0mUDLu
|
|
=Q3NN
|
|
-----END PGP SIGNATURE-----
|
|
|
|
|
|
</pre>
|