[ih] Fwd: History of "accounts"

[Jack Haverty]

In the spirit of capturing historical facts...

In the early 80s,  The Internet incorporated a satellite network, SATNET,
which provided connectivity between the US and Europe, with the
considerable expense absorbed by DARPA and associated sponsors.   Such
sponsors of course wanted that limited expensive resource to be used by
their projects.   There were other users who also wanted to communicate
between continents, but they weren't allowed (didn't pay for) use of SATNET.

I can't remember exactly when, but at some point while I was running
Internet projects at BBN, DARPA asked us to deploy a "VAN Gateway", which
was simply a gateway that used an X.25 network as a carrier, with an X.25
connection between 2 such gateways carrying IP traffic.   We did this, and
then there was connectivity between the US and Europe, using the X.25/X.75
public network as an underlying network instead of SATNET.   One gateway
was in the US, and one was in the UK.

I don't recall much of the detail, but I believe we relied on the two
European parts of the Internet remaining disjoint - no interconnections in
EU between the SATNET users' computers and the others'  - so that traffic
would go over the appropriate transatlantic path based simply by which
gateway (Satnet or VAN) a particular EU computer could access.   If they're
monitoring, Bob Hinden or Mike Brescia might remember more.

End-users didn't see any packet charges or other forms of accounting
feedback, but the US and UK government sponsors sure did, in the form of
bills from the X.25 providers.  Every connection, and every packet, cost
money, some in dollars, some in pounds.  We had to do some reworking of
internal mechanisms like gateway routing machinery, to avoid sending
packets constantly just because we could.

So, there was pressure to minimize operating costs by additional clever
techniques in the gateways.  The gateways could make decisions, for
example, about how long to keep a connection open, in case another packet
was heading its way to be sent out over the X.25 service.  As was typical
of the "phone system", each call made cost money, each packet sent cost
money, and each second the connection was open cost money.    Whoever
initiated a call got the bill for that call and for all traffic in both
directions.  Just like the phone system, pre-cellular at least.

So, ......

There was an interesting algorithm we put in the US-side VAN gateway which
wasn't discussed much and might have therefore escaped being captured by
History.   Until now.  Here it is, for posterity:

"When a packet arrives to be sent to the X.25 network, if the X.25
connection is open, queue the packet for sending.   If the X.25 connection
is *not* open, open the connection, send that single packet, and
immediately close the connection."

So, the SYN packet of a TCP connection heading to EU from a US computer
would get sent and the US would pay for a short, one-packet, X.25 session.
 The ACK returning from the EU computer would then open a connection from
the EU side, and, lacking a similar algorithm, subsequent packets for that
TCP session, and any others that might subsequently occur (FTP transfers
for example) would get billed to the UK.

So, most usage of that X.25 path got billed to the UK sponsors.  I don't
know if the UK ever noticed that this was happening.   Peter Kirstein might
remember.   For the curious, the history of the VAN gateway is discussed in
a series of BBN Quarterly Technical Reports of that era, many of which are
available online.

Someone told me once - "Management is the art of putting your expenses into
someone else's budget".... which we did, as part of our role in managing
the "core gateways".

Anyone thinking about Net Neutrality might want to look at what happened
back then.   It's far from a new idea for Internet carriers to treat
certain packets differently.

I hope I haven't started an international incident.....

/Jack Haverty
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20140212/40779a6d/attachment.htm>