Oct 21, 2005
Source code of the core compiled with GNU Java compiler (GCJ) see release 0.3.3c
May 09, 2005
Interop test with modem behind Ellacoya.
Sustainable 100KBytes/s up and 300KBytes/s down was reached
There are some good news and some bad news, but this is a major milestone and important day for the project.
The whole setup and test took under half an hour despite the fact that both particpating PCs were hidden behind firewalls and NATs.
It is fun to see something working
Feb 22, 2005
First beta. Contains data transfer and primitive chat
Feb 16, 2005
Lecture of Bram Cohen in Stanford. It was definitely worth driving from Santa Clara to Palo Alto. He is rather smart guy and very patient when asked
unrelated/stupid questions. He simply can not say f888 off i am tired and have to go NOW.
I succceeded to insert a couple questions of mine (i guess belonging to the same category of unrelated/stupid, etc.)
Uploader want to stop transaction at some point and smaller chunks give the uploader opportunity to
switch between peers frequently. When I started to argue that in case of 4Mbytes blocks it will take only 30s to transfer a block over reasonable connection
some guys chuckled. They forgot about 100Mbit connectors on their PC, that in Korea and Japan 10Mbit connections are not rare and that only five years ago 128Kbit
ISDN connections was considered superb.
- What was the reason behind small size of the data chunk (actually it's only 16Kbytes)
- We assume that the network is zero packet loss network and downstream is infinite. Is there anything wrong about argument that simply streaming the data
over IP is the most efficient way to deliver packets.
- Did he consider UDP based data exchange
Generally the answert is yes, but such networks do not exist. Couple of guys standing around even chuckled - yes, I was making a joke of myself all the way .
I decided to avoid discussion about how reliable the modern networks are and how more reliable they will get in the future when VoIP enters mainstream.
After all i am embedded telecom/datacom guy and see networks under different angle.
He considers using UDP between client and tracker, but he thinks that using UDP is an amateurish mistake. I have to agree with him, but there are
applications, where you simply do not have alternative. One of them is VoIP. When you use TCP it's like demonstrating your ID every time you cross street -
in reality it is even worse, because every time you have a small chat with policeman. There is no way to mask the address.
IP is not significantly better
but at least you can argue
that your IP address in connecitonless protocol can be faked by anybody. The major problem in connection oriented protocols like TCP that before you start
to send data underlying protocol runs non-trivial handshake over which you do not have any control. In the time (sometimes rather considerable time)
of this handshake your TCP/IP stack exchanges with remote TCP/IP stack all kind of information and it means that your IP source is specified correctly -
TCP establishes bidirectional connection. If you fake source IP in the IP packet remote side can not answer you and your TCP reports error to the application.
In other words you expose your home address to the whole world without even a chance to protect yourslef against mail spam. But, of course, this
amasing story is out of scope here.
The parametrizing of the problem of efficient data distribution is more complicated than I thought before, but I doubt that in the bitTorrent clients it was
solved in the best way and Bram expressed some doubts too. Still it really worries me that i missed in the design some major issues even after half an year of
work and constant improving.
Reliable network systems were metioned (this is what BitTorrent about) but the public was not prepared and did not catch the clue (?). Interesting issue here
is the number of simultaneous TCP connections regular tracker can establish (think about retransmission window) and number of TCP connections per second.
Theoretical limit is ~60K simultaneous connections - 16 bits of the port number, for every IP interface. In reality server with 2GB RAM can establish 30K connections
if we assume 64K window size. With larger round trip delays we get worse situation on the server. This is just another problem of TCP - it is not that scalable.
Feb 12, 2005
Icons for some of Rodi Houses are in place - Beta Testing, Big House, others
You can contact autohor of the icons Yuliya directly
and order graphic and WEB design works.
Feb 1, 2005
Rodi Houses are established - nod to some of my favorite fantasy books
Jan 30, 2005
All source can be found at Rodi CVS
Jan 20, 2005
Rodi release - Alpha
This is not beta, because there is no data transfer yet, but only LOOK (search) on the remote peer. It will take until end of Jan to
reach beta stage ( P.S. Feb 12 2005 ...sorry i failed to deliver on time as you see. the main reason is lack of design of data transfer)
Jan 18, 2005
Rodi has Project Manager
I can not name the person who accepted this position yet, but it is official from today we have project manager.
Beta release is delayed once again. Beta will happen hopefully Jan 20th.
Dec 08, 2005
I received some valuable feedbacks and decided to delay the beta release by about 2 weeks.
Some design changes are apparently required. Specifically search engines and data link protocol
Jan 5, 2005
The project is looking for beta testers. It is getting rather urgent.
Please send contact details to firstname.lastname@example.org
Dec 25, 2004
Rodi has icons
Dec 21, 2004
Run validator at http://validator.w3.org/ for the page http://larytet.sourceforge.net/btRat.shtml. It was interesting
exercise, but eventually i got
The document located at
was tentatively found to be Valid.
Dec 20, 2004
My goal of 200K of binary looks unreachable. Debug related CLI commands and
statistics cost about 50K of classes from about 150K existing (current JAR is 95K).
Java code is expensive in terms of binary size. I have to up the limitation to 0.5M. The
website will be updated accordingy. C/C++ implementation still can make it. Better arhiver
could do this too - i have 100+ small classes and this is worst case for JAR/ZIP
Dec 14, 2004
I am reading "Tarzan: A Peer-to-Peer Anonymizing Network Layer" paper. Many interesting ideas. thanks to www.i2p.net for the link
Tarzan's goal - to learn about all network resources - differs from recent peer-to-peer lookup protocols , which spend great effort
to achieve immediate information propagation and load balancing in a flat namespace, often at the cost of security ... node a simply contacts one
neighbor at random and transfers its entire neighbor set ... We select nodes by choosing randomly among the populated domains
at each level of the table in Figure 5. Tarzan uses a three-level hierarchy: first among all known /16 subnets, then among /24 subnets
belonging to this 16-bit address space, then among the relevant IP addresses ...
Dec 11, 2004
Sustainable packet rate 35K packets/s and byte rate 320 Kbyte/s. Currently i send only LOOK requests. Overall
performance is reasonable for Java applet. Available from release 18. Enjoy and do not use it to bring your ISP
Dec 05, 2004
William Jordan (see http://www.williamjordan.ca/) joined the project.
Welcome, William !
You can also visit http://home.bcwebnet.com/webdesign/ and
do not miss Josef Scott Design page and really great music at http://www.dennisburkemusic.com/db.html
Dec 04, 2004
I was sick a couple of days + i am moving to different appartments + i read "Effective Java: Programming Language Guide"
of Joshua Bloch. The book can help me to avoid most popular mistakes. Nothing signifcant changed in the current state
of the prject.
It is not clear what to do with memory pools. Any ideas ? I need pools of raw data blocks of different size
and pools of objects. Pool is implemented as a stack containing free objects. This gives very quick allocation/free
It is so easy with C++ templates, but i can't find any simple approach to the problem in Java.
The current implementation looks ugly. It is not clear also who needs this application when so many exist already.
In the last 3 months i received only one email and no comments from anybody. I see that people download the sources
from time to time, but that's it.
Nov 21, 2004
No frames version of the website is up. This is first step to XHTML. Loading of .shtml files is significantly slower. is it because of
preprocessing HTTP server does ?
Unrelated issue - this web site can be considered "Prior art".
Prior art is a legal term referring to whether an invention existed prior to the filing of a patent. All ideas
expressed on this web site are protected by GNU Public License. The author did not make any patent research and is not going to make
any patent research.
Nov 05, 2004
Email from potential user: My ISP just installed Ellacoya hardware BT blocking on all their NODES, looking into workarounds, any suggestions please???
Apparently the email was sent from client of
OrgName: Shaw Communications Inc.
Address: Suite 800
Address: 630 - 3rd Ave. SW
see also http://www.dslreports.com/shownews/56419
ISPs are getting smarter. I am wondering what they are going to do with HTTP based filesharing network ? And if the HTTP packet is
Oct 17 2004
we have debug CLI. It is limited and does not contain any usefull commands,
but the application is alive and running and it is under 60K of binary. Just another small step towards
alpha and then hopefully beta.
|Oct 9 2004
||we have logo. great thanks to Yuliya Abramova|
|Oct 03 2004
||PayPal button was added. It can be usefull to move money from one creadit card to another in cases when small amount of money spent.
Donations are greatly appreciated. According to Jason Rohrer 0.4% users donate money.|
|Sep 18 2004
||Arkady started blog
According to NedStat 17 September, 2004 18:36 someone with IP address belong to Cisco Systems, Inc., United States visited
|Sep 18 2004
||We use Firefox|
|Aug 23 2004
||P-Cube was bought by Cisco. Data encryption in P2P applications is our answer to this.|
|Aug 23 2004
||Actual coding started (download from http://sourceforge.net/projects/larytet/)|
|Aug 21 2004
||Name changed to Rodi (pomegranate in Greek)|
|Aug 19 2004
||btRat project was registered @ SourceForge as a separate project|
|Aug 16 2004
||Project returned from Tigris to the
original place - SourceForge (all issues related to upload files and CVS
were resolved). Frames added to the emLib home page|
Aug 14 2004
||Back to the problem definition and functional requirements. there are too many ideas. |
|Aug 13 2004
||Work with actual source code started. at this phase only framework is going to be created with some
sample application/engine for test purposes.|
|Jul 20 2004
||The project started @ Tigris.org as part of emLib project (in the future will be moved to separate project)|