[e2e] What should e2e protocols know about lower layers?

[David P. Reed]

At 09:19 AM 10/15/2001 -0500, John Kristoff wrote:
> > developers do this for a reason: they are generally frustrated
> > that TCP does not take full advantage of the capacity of a high
> > speed network, especially in the case of short connections. Make
> > no mistake, the pressure is going to increase with applications
> > such as ISCSI and gigabit networks.
>
>Perhaps developers should be encouraged to use another protocol in place
>of TCP, one that is TCP-friendly, rather than disable such a core
>function of TCP based on their own limited vision of the network?

Perhaps a sarcastic remark is in order.  My hard drive has a 100 MHz bus, 
but I am incredibly, sadly, ragefully, personally disappointed that my 
applications don't get "full use" of it.   I don't know why all Mac, Sun, 
PC users don't rise up and demand their money back! I suppose I should get 
really upset and start a class action suit against all the software and 
hardware vendors involved...

Seriously, taking seriously the idea that ANY bus or link should be "fully 
utilized" calls for direct application of the clue-by-4.  Does anyone 
really think that because a car can go zero-to-60 in 7 seconds that that is 
what should happen at every light-change?  Or that cars should always 
corner at the maximum G-force offered by their suspension and tires?

The point of high capacity local buses and networks is to create unused 
capacity to reduce latency, queueing, and congestion overhead.

>John

- David
--------------------------------------------
WWW Page: http://www.reed.com/dpr.html