[TI-H] Re: TI Networking.
[Prev][Next][Index][Thread]
[TI-H] Re: TI Networking.
Quoting jeff <jefromi42@yahoo.com>:
> I really know nothing about network protocols, but it seems like this
> must have come up when
> people were figuring out how to network computers. How do real networks
> between computers keep
> from having collisions and stuff like that?
If you are talking ethernet, they don\'t. Ethernet cards just wait until
the line appears clear, then they start talking. If the card detects a
packet collision, it stops talking, waits for a random delay, then starts
over. In very high traffic networks, the data rate can go way down
because of high collision rates.
In a token ring (thats logical ring, not necessarily physical), each
device keeps quiet until it has the \'token\', a logical construct that
is passed around. Only the device with the token is allowed to talk.
Collisions on a token ring network are very rare, and the amount of
traffic does not impact the data rate as dramaticly as in ethernet.
Depending on the physical setup, token rings can do some cool
beaconing to determine where breaks in the network occur, and
to operate in spite of such breaks.
There are other protocols too of course, like the master-slave
arrangment that I2C uses.
> Is there any way that a real protocol could be
> \"ported\", or do they all require something more than just
> three wires that go to each unit? It seems like sometime
> this problem must have been solved sometime before.
The main problem is balancing ease of use, development time
and hardware complexity.
I personally would build a hub that acts an an I2C master
and passes data between calcs, which act as I2C slaves. I
think this solves almost all of the problems, including
address assignment, as the hub can act as a sort of I2C
DHCP server, assigning new addresses from its pool when
a calc gets on the network.
DK
Follow-Ups:
References: