Why has this not been done?

[dave walker]

Niall Sheridan wrote:

>dave walker wrote:
>  
>
>>So when drinking some beer with friends we sometimes come up with crazy
>>ideas that should be done, however we normally figure out why it is a
>>bad idea.  However, I have had an idea stuck in my mind for about 2-3
>>months and can not figure out why it has not yet been done.  I know
>>there is randomness in the initial sequence number for TCP packets.  But
>>what about making the rest of the packets random sequence.  (read my
>>post before saying you can't have random sequence numbers)
>>    
>>
>
>There's a good reason to have consecutive numbers. It's called the "I
>can tell which packets I have received and which packets I have not" reason.
>  
>
>>I know there is a lot of talk going about with network security and all,
>>and I am not sure if it is relevant to this idea that I have.  I know
>>most talks of security are with phishing and what have you.  I thought
>>of a hack for TCP that would deter man-in-the-middle / hijacking
>>attacks.  I wrote a small thing on it here:
>>http://blogbp.com/archives/2005/07/security_throug.php
>>    
>>
>
>And how exactly do you tell the far end what equation to use without it
>being subject to a MitM attack?
>  
>
>>The only problem we see with it, is the potential for it taking time to
>>come up with (int)1 to (int)1 equations.  So if there is a better reason
>>on why this has never been done I would love to know.  I am not the best
>>at TCP Protocol programming, and don't really understand what happens at
>>that level.
>>Thanks,
>>    
>>
>
>It has never been done because it has the unfortunate side effect of
>killing TCP in the face.
>
>~ Niall
>
>  
>
If I miss the handshake I would have to figure out how the numbers would 
be sequential.  If I got the handshake I would have to break PGP, which 
yes is subject to a MitM attack, but we are subject to far less packets 
vulnerable to MitM over all.  Yes, it does hat that nasty side effect of 
killing TCP in the  face, but if there is an easy way to make sure an 
equation passes the vertical line test, and has enough randomness 
within, it may be easy to get the 1 to 1 equation to use. 

Something like what I am talking about might do better up on the OSI 
stack, as something that can be enabled by software.  And TCP for that 
session would enable the randomness or something.

--
Dave