[TI-H] Re: TI Networking.

To: ti-hardware@lists.ticalc.org
Subject: [TI-H] Re: TI Networking.
From: Olle Hedman <oh@hem.passagen.se>
Date: Sat, 05 May 2001 20:41:01 +0200
In-Reply-To: <20010505180041.62F222747@sitemail.everyone.net>
Reply-To: ti-hardware@lists.ticalc.org

At 20:00 2001-05-05, you wrote:
>Then don't send the whole thing. Hash the 56 bits into a checksum or other 
>nifty byte/nibble format that will retain the uniqueness of the Calc. 
>Exactly how without encountering a duplicate at least once, I dunno. 
>(Maybe increment if a duplicate is found?) But starting with the Calc's ID 
>number is a great idea. If the network hub is only about 4 ports or 8, you 
>can actually hash it down to 2-3 bits and increment when collisions occur 
>(one will get the address before another) and actually put the hub into 
>the state of a basic router sending packets to the right addresses, which 
>is practically required with the TI's small bandwidth potential.

This assumes an intelligent hub. Doesn't go with the simple-hub approach 
that is.
Ofcourse that is nothing wrong. I though will concentrate on the simple-hub 
approach for a while.
It is my belive that the simpler the "hub" is to build, the more people 
will use it.
If you are going to check for adress collisions, you can just as well 
assign incremental adresses from the beginning, and totally ignore the ID.
going from x bits to y bits always produce duplicats when y < x. That is 
kindof obvious :)
Having adresses that isn't next to each other doesn't really work well with 
the turn based shared media model suggested here either.

///Olle

References:

[TI-H] Re: TI Networking.

From: Adam Thayer <krevinek@matrix-z.net>