[GNUnet-developers] MySQL/Service/three structure/gnunet.conf

[Jan Marco Alkema]

Hi all,

The current CVS gnuetd version uses less memory ---)

I see in mine MySQL database (with fshpeerinfo which start up every 5
minutes in cron) that 55 different Gnunet servers have operated in the time
interval (20 Februari 2003, now).

In mine opinion it is very important to have a "driver" to spread Gnunet on
the Internet. A driver is a program/aspect that is very useful for the user.
A driver could be content (Mp3, DVD, CD, etc) or a very simple manner to
download all the files from:
http://cerealport.homeip.net/cereal/PDNJAVA/filesbbs.html If I want to
download them with the explorer it cost me a lot of time.

Cerealport.homeip.net could make the files in a service application, with
comes only online when you want to known which files the host has on the
computer for sharing or when you want to download them.

Christian, I known that gnunetd initial logs in to other Gnunet servers with
the list from  http://www.ovmj.org/GNUnet/download/hostlist . I see that ftp
(vsftpd) uses the Linux service facility.  Suggestion. Make Gnunet also
available as Linux service. Default no networkload should not generated
extra network load by the service. Maybe other port (2086) could be used. If
someone logs in the server the Gnunet service comes up. If that person logs
off the service goes down. A known public port (like 21 for ftp) is very
useful. People can make an IP scan to discover all the Gnunetd server on the
Internet.

Igor, I am very happy to see that a port to MySQL will be made in
Gnunet ---)

>As usual, I have no idea what you're talking about :) Do whatever you feel
is necessary to make yourself happy.

Krista in mine view there are 3 parts of software:

1) {Java GUI + jdbc} for Linux/Windows. Windows version should have a
installer program. 12 march 2003 the new Cdfoon 2003 will be out --). For
the foonrip project I have to make a port to JDBC. Because of the
intellectual property problem I will name it foonsearch.jar.
2) Windows GUI (Microsoft Visual C++ (Studio) compiler) for Windows. This
version should have a Windows installer. I will try rewrite existing Windows
GUI source code. I begin with www.grub.org.
3) {gnunetd + vsftpd + odbc} for Linux and Windows (Windows could have two
ports. One is Cygwin and the second is Microsoft Visual C++ (Studio)
compiler). Optimal will be that the three compilers (GCC, Cygwin, Visual
C++) can build the core engine.

> The party line is that if the node hasn't satiated its
connections (the limit specified in gnunet.conf), it will
aggressively try to establish new connections. The
connection table is impossible to fill if the limit
is 64 (as default) and yet there's only 10 or so nodes
online concurrently. And if there's no other traffic,
the node will still forward HELOs frequently, perhaps
it'll change session keys too (not sure), but always
the difference between the content size and MTU will be
padded with noise, or if enough space is left, with
low-priority data.

I don't think this is a good thing I believe more in the "three structure".
If I look now to the picture of tracekit in the documentation I see a not
structured way of links between the core engines. Servers/internet links
with high capacity should be in biggest part of the three. N.B. A coffee
machine on IP should be far from the middle. There should not be tons of non
info send to the coffee IP machine.

A core engine get an IP address from the root (www.rootserver.com). The
client gives info about internet link capacity and the service he has to
offer (time server, catalog of lot of dvd/cd/programs, availability of 24
hours a day 365 days a year, etc.).  If there are to much core engines
logged in the rootserver then the rootserver asks an appropriate core server
to operate like the rootserver. In the login procedure a core engine can
tell that he has the ability to do so. Core engines can politely been asked
to go to the new rootserver. New incoming core engines will automatically be
moved ("rerouted") to the new rootserver. The core engine can be moved a lot
of times to other rootservers. If the link from core engine to the three is
broken then the core engine will start at the root of the three. He tries to
login  www.rootserver.com.

The manner of moving to a certain rootserver can be policed on the
communication need of the users of the network. All MP3 users in a certain
virtual domain. All "adult video" users in another domain. A domain is a
part of the three with high correlation of needs of the users of the core
engines

>- too many options in gnunet.conf, remove "bad" ones (that are not strictly
needed)

In mine opinion /etc/gnunet.conf should have the general configuration
items. In a database should be the specific configuration items like max
network load. If you "configure" gnunet you insert records in a
configuration database. For instance the system can calculate that the max
network load should be downgraded. In the database you can see when (time
stamp) the max network load for gnunetd is downgraded. If you put
configuration items in a database you can easely have multiple gnunetd
processes on one Linux server. Remote support is also an issue.

If someone has better ideas, please let me known ---),

Greetings Jan Marco