[ih] Fwd: History of "accounts"

[Brian E Carpenter]

On 13/02/2014 00:46, John Day wrote:
> Actually, it was quite common at the time for networks to charge for
> both connect time and bytes sent.  Datapac, Transpac, Telenet, Tymnet
> all did this.  

It was a clear design advantage of X.25 for the PTTs that it made
such charging much easier. There was even a theft of service attack
possible at one stage, when they *didn't* charge for the initial
connect packet, which contained a Call User Data field that the user
could set to any value - after a while they started charging for
failed connects, so that people couldn't send free bits in that field.

> One of the arguments against datagrams at the time was,
> how to charge if the network had no idea what packets were being
> delivered and which ones weren't.  Even they realized that they
> shouldn't charge for retransmissions.

I wasn't in New Zealand at the relevant time, but I understand
that charging by the Internet byte was normal due to the high cost
and scarcity of trans-Pacific capacity in the 1990s.

   Brian

> 
> There seems to be a lot of conjecture going on here.  Generally, when we
> do history we try to consult original sources to document the facts as
> much as possible, rather than simply conjecture what sounds reasonable.
> 
> John
> 
> At 4:57 AM +0000 2/12/14, John Levine wrote:
>>  >I don't think the absence of a bean-counting-bit impeded the growth of
>>> the internet, or the remuneration of those who built the infrastructure
>>> to enable it. In fact, I'd say that *at the level of IP and TCP*, the
>>> ideology of the shared-information sorts actually enabled a network
>>> that could be billed at connection- and application-levels.
>>
>> That certainly seems right to me.  Trying to account a packet at a
>> time would have been a huge distraction from getting useful work done.
> 
>