[ih] message vs. packet

[brian.e.carpenter at gmail.com (Brian E Carpenter)]

On 08/07/2018 12:39, John Levine wrote:
> In article <CAHxHggf7L0qH5o-B6-kkKJxVxH6VfYh6bVcRrgJ3cJJSeT6aeA at mail.gmail.com> you write:
>> The basic methods of queueing theory apply to both message and packet
>> switching.
> 
> Seems to me the key insight is that you can number the packets and the
> recipient can reorder them so the network doesn't have to worry about
> keeping them in order.

Maybe, but ATM went for small fixed size packets to get to what they thought
was the sweet spot in queueing theory**, but they had to stay strictly in order.
And anybody who's written reassembly code knows what a bad thing out-of-order
variable-length packets can be.

** and the queueing theory is definitely simpler with fixed size packets,
because one of the distributions becomes a constant. In my misspent youth,
I slogged through both volumes of Kleinrock's book. I don't remember much,
but I do remember that M/D/1 is a lot simpler than M/M/1.

    Brian