jared and daniel,

What advantages does Ttorent has:

1. Torrent itself is a protocl. It should be possible to implement it even in
   Perl

2. In Case of Universities wich have maybe NO mirror:

Some universitives have a VERY liberal administration.
In Fact that means that every part of the univeristy (math department,
computer science, languages, history and co) does administrate it`s OWN
LAN by itself.

There`s no rule that e.g. a small Network must be managed by an
"Administrator". So every doctor who`s responsiable for a little Lan can
do what he thinks is right.

Some universities have Proxies. But these Proxies do mostly NOT buffer
such large files. And some even do not download all the stuff used int he
university and store it at a central Server (in their Lan).


This leads to multiple downloads and so they need to download e.g.
packages and co more then once.

With torrent they would be forced to use also the "uploads" from their Lan
if somebody downloads e.g. the Install-Sets for Sparc and another
department downloads it too. This would reduce the load of the Servers.

2. Small Files and Big files

I did never thought about forcing users to "share".
They should simply just share as long as they download. After the packet
was downloaded they stop sharing.

What does this mean: Servers may have to download small files completly to
a client. There4s no differente to FTP here.
But downloading BIG-Files is different. There the download takes much
longer and so the client are able to upload also much longer.

3. Security

You trust e.g. Beck because he`s known.
OpenBSD wont change the compression-algorithm but imported gzsig to the
Base-System. Why not simply using gzsig to SIGN the files.
Then it wont matter who provides the data because the data was signed and
the Port-System would simply just to maybe look for new Keys (if Theo
changes the SSH-key some day).


It is importent to understand that I DON`T talk about a normal BT-Client
wich simply seeds all the day long until it get killed.
The Client may seed as LONG as it downloads. After the download finished
it stops.

Where is the advantage: Even a Client just seeds 1/8 of the original file
this means the Server has 1/8 less traffic.
The more Servers share Seeding == the more synergy-effect you get.
So it wont happen that some Servers may have a lot more traffic then
others just because they`re maybe the only one in e.g. romania (just as
example) or because they`re maybe prefered by a lot more users.

During the upload of the Clients during downloading and during the seeding
or maybe more then just ONE Server the bandwith needed to transfere filles
gets down (for everybody because it`s imply shared by more then one Box).

Sure FTP and HTTP works but the bandwith has to be paid by somebody.
It is nice that some peoples or companies/universities do support OpenBSD
but why not inventing a System wich may reduce even their coasts.

That`s the whole Idea behind my post. Nohhing more or less.
it was not because I love Torrents that much but indeed it seams to be the
best protocol for this purpose.

It has also already an advantage where e.g. 2 Administrators (lets say one
in USA and one in Europe) of the same company may do a Firewall update or
install an openBSD-Box. They don4t have to knwo that the other Admins is
doing also e.g. an Update in the weekly update-circle because
Torrent-Would notice it and may send data about e.g. the VPN-Tunnel (so
the LAN) to the other guy.

This example assumes that BT-Traffic isn`t blocked in the LAN but for
Universities wich e.g. use NAT or simply don`t use Proxy-Servers this
would already reduce the load effectivly.
I don`t think the Servers can save 50% of the Traffic but 10-30 percent
seams to be realistic. As mentioned above the "real" advantage would
appear if a Client installs a big package.

I even bet a lot peoples simply try out OpenBSD after reading e.g. an
article (like on kerneltrap). So imagine just 100 would download also the
ports.tar.gz and the source-code...

The current system "works" but it isn`t as effective it may could be.
And I wanted to use Torrents today to get the AMD-Packages but as I
re-visited the Torrent-Website not even the Server 8Tracker) itself shared
the packages. This means, if I would work for a company and would be a
leats a littlebit clever, I would download all binary packages to a local
Server to spread them in the LAN (if needed). This means 2.x Gb traffic
for ONE Server. If there would eb ANY organized Torrent-Network the load
of the Server would be maybe just 1.8GB because somebody else downloads
the packages too and would upload in the Same time.

That`s the whole idea... and "not" powering Python (even the License seams
to be BSD-like and smaler *compared to Perl*).

Kind regards,
Sebastian

Reply via email to