Re: New idea in TCP

[vint cerf <vcerf at MCI.NET>]

there is the problem of spoofing that makes this somewhat
more difficult that is outlined below.

vint

At 02:43 PM 4/2/2001 +0800, R.Z. Pan wrote:
>Hi, all:
>  I'm a newbie from P.R.China and I do some load balance stuffs. I noticed
>that HTTP Protocol has a redirection mechanism to lead user to another web
>site. Why TCP Protocol don't use this redirection mechanism to implement
>load balancing? In my opinion, the process would be like:
>                   SYN
>User           ------------>       Load Balancer
>
>                Redirection
>Load Balancer  ------------>       User
>               Best Server's IP
>
>                   SYN
>User          ------------->       Best Server
>
>                   SYN
>Best Server   ------------->       User
>            ACK
>
>                   ACK
>User         -------------->       Best Server
>
>
>The "Best Server" is one of the load server that be choiced by the Load
>Balancer
>
>To implement this, TCP Protocol should add one "redirect" bits. Data with a
>redirection control bit we called it a redirection data, the redirection IP
>address packed in the "Options" section of the TCP Header. For the TCP data
>bring us the port number, so we can also dispatch the job according to the
>port. I think load balancing can be resolved by the Operation System in this
>way.
>Is this a nonsense or a good idea? Please give me a advice. Thanks for your
>attention!