New $Support: SmartProtcol

Corayzon · Post by **Corayzon** » 2006-01-07 09:15

Hey guys,

Ive been working with dc for quite some time now, and ive come up with an idea that will allow hubs to contain at least 2 times more users

I've called it SmartProtocol. Here's the jist of it:

Instead of the server taking the load for bouncing $Search's and $ConnectToMe's around, SmartProtocol will allow clients to do the work.

It involves clients sending their tcp\udp info in the $MyINFO string

- $IpAddress:tcpPort:udpPort$

This means when users connect to the hub, the hub will send the tcp\udp info for each user to each user!

Now that each user in a hub has eachothers Address and Ports. They are able freely search and connect with eachother without interacting in the hub.

Now this does have one exploit if left open. It makes all the users open in the hub. So each user MUST DISCONNECT any connecting user on the tcp port when the $MyNick and IP Address doesnt correspond to a user connected in one of the connected hubs

I would also suggest adding a $HubName command so users can check what hub the connection is comming from.

This also has one more very good side to it.

IT STOPS WANKERS BEING ABLE TO FLOOD DC USERS TO PREFORM DOS ATTACKS!

So i ask, What do we think?

FarCry · Post by **FarCry** » 2006-01-07 09:55

ADC offers that and the risks have already been discussed here.

ivulfusbar · Post by **ivulfusbar** » 2006-01-07 10:26

$HubName has never been static. Different users can recieve different HubNames while being connected to the same hub. So hubname can not be trusted or used in this context.

I can see several more flaws in you suggestion but i urge you to read up on the ADC-proposal farcry pointed you towards before adressing them.

Corayzon · Post by **Corayzon** » 2006-01-07 10:32

FarCry wrote:ADC offers that and the risks have already been discussed here.

ADC is a new protocol and isnt based on the current dc protocol whatsoever.

Im not to sure on it all,but if u have 2 types of protocols in the one hubsoft, the hub then has to translate protocol calls for each client.

This just wont happen in my mind

What im talking about is an extension of the current protocol so it can support more users and take out the exploits that cause dos attacks forever!

Corayzon · Post by **Corayzon** » 2006-01-07 10:39

ivulfusbar wrote:$HubName has never been static. Different users can recieve different HubNames while being connected to the same hub. So hubname can not be trusted or used in this context.

I can see several more flaws in you suggestion but i urge you to read up on the ADC-proposal farcry pointed you towards before adressing them.

Well, im not sure u got me on the $HubName protocol command.

In my idea, this is how it would work as a short protocol example:...

Client1 connects to server and sends tcp\udp info with $MyINFO
Client1 sends its $Search to all clients on their upd ports (without server)
Clients reponse to the $Search as per normal on the udp port (yes this can be exploited,but no worse then it currently is)
Client1 Connects to Client2 for download\upload
Client1 Sends $MyNick $HubName
Client2 Checks if $MyNick is in $HubName with the correct IP Address
Client2 Disconnects Client1 when fakes a connection attempt for downloading\uploading

It's flawless, and fake clients cant get into the hubs clients unless they are connected to the hub!

Futher discussion is needed =] ...

[NL]Pur · Post by **[NL]Pur** » 2006-01-07 10:46

Client1 sends its $Search to all clients on their upd ports (without server)

How do you verify the identity of client1 in case of the second step sending the search ?

Corayzon · Post by **Corayzon** » 2006-01-07 10:49

[NL]Pur wrote:Client1 sends its $Search to all clients on their upd ports (without server)

How do you verify the identity of client1 in case of the second step sending the search ?

hmm, yea ur right,

Well i guess we would need $MyNick and $HubName in the search, as it already contains the ip address.

Then the clients can dismis the $Search command if its fake (from a kazza network! ewww)...

[NL]Pur · Post by **[NL]Pur** » 2006-01-07 10:51

Bare in mind that:
it's udp, ip are easy fakeable many clame that
you know with your construction everyones nick+ip combination

Corayzon · Post by **Corayzon** » 2006-01-07 10:52

[NL]Pur wrote:Bare in mind that:
it's udp, ip are easy fakeable many clame that
you know with your construction everyones nick+ip combination

but why would u fake the ip in the $Search string?

Because no clients would be able to send the search returns.

So it would defeat the purpose of trying to search

[NL]Pur · Post by **[NL]Pur** » 2006-01-07 10:53

if i spam everyone in the hub with udp searches using your nick+ip, you won't be happy receiving the results.

FarCry · Post by **FarCry** » 2006-01-07 10:54

Corayzon wrote:ADC is a new protocol and isnt based on the current dc protocol whatsoever.

The problems are the same for both protocols. In ADC 0.10, UDP searches will be restricted to trusted environments, because of the dangers brought up in that wiki article. Your goal was to increase the number of users a hub can handle, which is seldom an issue for the small communities someone considers trusted. For an extension to a protocol that already has numerous client and server implementations running, the possible gain is not worth the effort in my opinion. I personally doubt its value even for ADC.

Corayzon · Post by **Corayzon** » 2006-01-07 10:54

[NL]Pur wrote:if i spam everyone in the hub with udp searches using your nick+ip, you won't be happy receiving the results.

This can allready be done in the current protocol, because not many hubs check if the ip address's match

[NL]Pur · Post by **[NL]Pur** » 2006-01-07 10:57

But that would be something of wrong implementation and not flawed by design. But correct me if i'm wrong there.

Corayzon · Post by **Corayzon** » 2006-01-07 10:59

[NL]Pur wrote:But that would be something of wrong implementation and not flawed by design. But correct me if i'm wrong there.

Yea ur right,

But then at least these dos attacks can only take place on other users within the hub, rather then any tcp\upd connection in the world.

The risks are far less then the current protocol, correct me if im wrong ^^

Corayzon · Post by **Corayzon** » 2006-01-07 11:01

FarCry wrote:
Corayzon wrote:ADC is a new protocol and isnt based on the current dc protocol whatsoever.
The problems are the same for both protocols. In ADC 0.10, UDP searches will be restricted to trusted environments, because of the dangers brought up in that wiki article. Your goal was to increase the number of users a hub can handle, which is seldom an issue for the small communities someone considers trusted. For an extension to a protocol that already has numerous client and server implementations running, the possible gain is not worth the effort in my opinion. I personally doubt its value even for ADC.

Ive read this like 5 times, and it just keeps going out the other ear :S

Im not sure what ur pointing towards there sorry FarCry

[NL]Pur · Post by **[NL]Pur** » 2006-01-07 11:02

I'm not sure now to what kind of weakness of nmdc you are refering too

Corayzon · Post by **Corayzon** » 2006-01-07 11:05

Hmm, yea, any current hubsoft can correct the issues with dos attacking with only a few lines code.

But remember, im not trying to fix any current issues. Im only trying to make this network bigger, better and smarter.

Im allmost sure if we sat down and thought about the $Search exploit in my implementation, we could devise a solution that removes the risks

FarCry · Post by **FarCry** » 2006-01-07 11:10

Corayzon wrote:Im not sure what ur pointing towards there sorry FarCry

Just like Pur explained to you (and the wiki article describes in detail), your idea comes with a vulnerability to client and hub DDoS attacks, which makes it only usable in environments where every client trusts every other client. You should perhaps have followed the link instead of just saying that it's a different protocol.

FarCry · Post by **FarCry** » 2006-01-07 11:13

Corayzon wrote:Im allmost sure if we sat down and thought about the $Search exploit in my implementation, we could devise a solution that removes the risks

Others sat down about this and the only possible solution seems to be signing those searches to avoid identities being spoofed, which is inacceptably complicated.

Corayzon · Post by **Corayzon** » 2006-01-07 11:15

FarCry wrote:
Corayzon wrote:Im not sure what ur pointing towards there sorry FarCry
Just like Pur explained to you (and the wiki article describes in detail), your idea comes with a vulnerability to client and hub DDoS attacks, which makes it only usable in environments where every client trusts every other client. You should perhaps have followed the link instead of just saying that it's a different protocol.

silly me not reading shit properly,

it will be the death of me i tellz ya

Corayzon · Post by **Corayzon** » 2006-01-07 11:24

Damb, drops head in shame

Todi · Post by **Todi** » 2006-01-07 14:42

Corayzon wrote:Damb, drops head in shame

You should have done that already when you named your extension "SmartProtcol"

(although, i guess it's kinda in the spirit of nmdc spelling)

Snooze · Post by **Snooze** » 2006-01-07 16:43

Real helpfull comment, Todi..

sandos · Post by **sandos** » 2006-01-08 18:59

Snooze wrote:Real helpfull comment, Todi..

Real helpful comment, Snooze..

Corayzon · Post by **Corayzon** » 2006-01-08 21:34

this could go on for ever,

MagicShroom · Post by **MagicShroom** » 2006-01-10 13:35

And it will so long as everyone reads it.

Ah well, it's some times good to descuss these sort of things.