Hornet
December 11th, 2004, 01:16 PM
The quotes below are taken from what I remember "hearing" on Ants IRC:
Ants and IRC
Recent discussion between two Ants users on IRC about IRC,
“IRC is ANts biggest security hole.”
“It must be closed and used for bootstrapping only.”
“Look at the Winny case.”
“Yes, closed except for bootstrapping and random connections”
“All that needs to be done is to have chat over Ants. All it needs is just a GUI, nothing more.”
“Ants can send text messages already to a DOS window if you know how. Gwren said all that is needed is a GUI.”
So what do you think, should Ants keep chat on its built in IRC channel or should an anonymous chat be developed?
Personally, I enjoy chatting on Ants but an anonymous channel would be better as then you could request files and tell other users what files you’ve got.
Bit Torrent Plug-in
This was what was said about Bit Torrent by Gwren in conversation with an Ants user
“Gwren it’s time to update 'Bit Torrent plug-in'?”
” Whoever mentioned a Bit Torrent plug-in?”
“You did in your post, http://sourceforge.net/forum/forum.php?thread_id=1189812&forum_id=37761.”
That is not a plug-in. It is simply an idea, probably impossible to realize because of anonymity issues.
Anyway, talking about it, is not doing it. I simply asked to the Azureus people if they have any idea on how to make it possible. The problem with the plug-in is the loss of anonymity.
I'll explain briefly, Bit Torrent works like this. You have a torrent file that identifies a tracker. This is the core idea. The other stuff is not relevant.
The tracker is made up of an http server and a relative IP address. To use torrent files you NEED a unique and CONSTANT id.
But in Ants this is what you will never be able to have. If you have a unique id that is always the same then it could be easily linked to your IP.
Ants uses a technology "similar" to bit Torrent. I'm thinking about a way to integrate the systems but there are two issues that are not easily solved:
1) Tracker identification and
2) Linking networks
Ants is not Freenet. In Freenet you have a "distributed website" with a unique id. That free site is spread all around the net but is unique.
In Ants you have a web site that changes its identification number every time you log on. It is something like a dynamic IP on the internet. You know with a dynamic IP you cannot be a tracker or you can be a tracker only for a while.
So the only solution that comes to my mind is this, performing queries to find a suitable tracker for your torrent file. You can apply Bit Torrent technology INSIDE Ants.
You cannot join anonymous and NOT anonymous net. This is the basic rule.
What I was thinking about was integrating Bit Torrent technology INSIDE Ants. That could be done but there is a problem.
A tracker is unique - this is the only problem. I don't need a unique tracker. I need many equally likely trackers, just like hashes, so any node could be the source provided it has the file.
If this can be done, we are done because a tracker can be identified through queries and the next steps can be performed without problems. So it would be the same as a hash in simple terms, just with more information.
However, I'm wondering if that could actually give better performance. Bit Torrent is fast because it spreads chucks in a tree-like way in Ants that is not possible.
You have always to route through other nodes. So we have to forget about the tree and probably have "logic trees" that will be "fit" inside graphs. The current Ants structure is cyclic graphs. So I'm actually thinking about the real complexity of this stuff.
However, probably with the given constraints, Ants is one of the best solutions to the zero knowledge routing. We have to think about tools and technology we have, not about science fiction.
So basically, probably a waste of time to implement Bit Torrent in Ants or to link Ants to Bit Torrent.
Download Speed of Ants Network
Gwren also had some thoughts on the speed of the Ants network.
“Roher found a nice way to anonymize file sharing. The problem is that this solution has a high cost. Now the question is it possible to lower this cost or have we already reached the minimum? In my opinion Ants is very close to the minimum for this problem. That is, in my opinion, a 20% increase of Ants performance is to be excluded with the current net structure. With a broadband connection (not ADSL) the performance problems will be overcome.
Anyway I don't know how many nodes MUTE has, probably it has 40/50 nodes but it is actually really slow, much slower than Ants. I tried today to download a simple mp3 and everything stalled after a few minutes so Ants is not so bad.
So with current network topology download speed probably won’t increase by very much.
Currently Ants is faster than MUTE.
Hornet
Ants and IRC
Recent discussion between two Ants users on IRC about IRC,
“IRC is ANts biggest security hole.”
“It must be closed and used for bootstrapping only.”
“Look at the Winny case.”
“Yes, closed except for bootstrapping and random connections”
“All that needs to be done is to have chat over Ants. All it needs is just a GUI, nothing more.”
“Ants can send text messages already to a DOS window if you know how. Gwren said all that is needed is a GUI.”
So what do you think, should Ants keep chat on its built in IRC channel or should an anonymous chat be developed?
Personally, I enjoy chatting on Ants but an anonymous channel would be better as then you could request files and tell other users what files you’ve got.
Bit Torrent Plug-in
This was what was said about Bit Torrent by Gwren in conversation with an Ants user
“Gwren it’s time to update 'Bit Torrent plug-in'?”
” Whoever mentioned a Bit Torrent plug-in?”
“You did in your post, http://sourceforge.net/forum/forum.php?thread_id=1189812&forum_id=37761.”
That is not a plug-in. It is simply an idea, probably impossible to realize because of anonymity issues.
Anyway, talking about it, is not doing it. I simply asked to the Azureus people if they have any idea on how to make it possible. The problem with the plug-in is the loss of anonymity.
I'll explain briefly, Bit Torrent works like this. You have a torrent file that identifies a tracker. This is the core idea. The other stuff is not relevant.
The tracker is made up of an http server and a relative IP address. To use torrent files you NEED a unique and CONSTANT id.
But in Ants this is what you will never be able to have. If you have a unique id that is always the same then it could be easily linked to your IP.
Ants uses a technology "similar" to bit Torrent. I'm thinking about a way to integrate the systems but there are two issues that are not easily solved:
1) Tracker identification and
2) Linking networks
Ants is not Freenet. In Freenet you have a "distributed website" with a unique id. That free site is spread all around the net but is unique.
In Ants you have a web site that changes its identification number every time you log on. It is something like a dynamic IP on the internet. You know with a dynamic IP you cannot be a tracker or you can be a tracker only for a while.
So the only solution that comes to my mind is this, performing queries to find a suitable tracker for your torrent file. You can apply Bit Torrent technology INSIDE Ants.
You cannot join anonymous and NOT anonymous net. This is the basic rule.
What I was thinking about was integrating Bit Torrent technology INSIDE Ants. That could be done but there is a problem.
A tracker is unique - this is the only problem. I don't need a unique tracker. I need many equally likely trackers, just like hashes, so any node could be the source provided it has the file.
If this can be done, we are done because a tracker can be identified through queries and the next steps can be performed without problems. So it would be the same as a hash in simple terms, just with more information.
However, I'm wondering if that could actually give better performance. Bit Torrent is fast because it spreads chucks in a tree-like way in Ants that is not possible.
You have always to route through other nodes. So we have to forget about the tree and probably have "logic trees" that will be "fit" inside graphs. The current Ants structure is cyclic graphs. So I'm actually thinking about the real complexity of this stuff.
However, probably with the given constraints, Ants is one of the best solutions to the zero knowledge routing. We have to think about tools and technology we have, not about science fiction.
So basically, probably a waste of time to implement Bit Torrent in Ants or to link Ants to Bit Torrent.
Download Speed of Ants Network
Gwren also had some thoughts on the speed of the Ants network.
“Roher found a nice way to anonymize file sharing. The problem is that this solution has a high cost. Now the question is it possible to lower this cost or have we already reached the minimum? In my opinion Ants is very close to the minimum for this problem. That is, in my opinion, a 20% increase of Ants performance is to be excluded with the current net structure. With a broadband connection (not ADSL) the performance problems will be overcome.
Anyway I don't know how many nodes MUTE has, probably it has 40/50 nodes but it is actually really slow, much slower than Ants. I tried today to download a simple mp3 and everything stalled after a few minutes so Ants is not so bad.
So with current network topology download speed probably won’t increase by very much.
Currently Ants is faster than MUTE.
Hornet