[ih] Anticipating network and internetwork scaling

[John Shoch]

In the last week there have been a number of interesting comments on
stateless gateways, which led to a broader question -- if one could have
predicted the eventual scaling of networks and the internet.

On Jan. 26 jnc wrote:
"I don't think we could have been thinking 'this aspect of lack of state in
the internet packet switches [m]eans it will scale indefinitely' because I
don't think we had any idea, at that point, about how to do path selection
in
a global-scale internet - so global-scale internets could not have been in
our thinking."

On Jan. 26 Vint Cerf responded, in part:
"I don't know that we recognized the scalability aspect

but we definitely cared a lot about statelessness of the gateways"

On Jan. 28 Len Kleinrock commented on the scalability of packet networks.

On Jan. 28 Brian Carpenter noted:
"I just glanced through Baran's 1964 paper, and it clearly recognized
statelessness (and a standard packet header) as important for network
survivability and adaptive routing. But although he mentions networks
of intercontinental size, I didn't spot any discussion of scalability
as such.
Interestingly, exactly the same applies to Dave Clark's 1988 "Design
Philosophy" paper.
In RFC 1958 [published in 1996], we did note as principle 3.3 that "All
designs must scale readily to very many nodes per site and to many millions
of sites".
I guess that by then (1996) this was too obvious to ignore, and it was
written when IPv4 address exhaustion was considered inevitable.
Maybe somebody who knows the early literature better than me can find
something. But it's almost as if the intrinsic scalability of stateless
packet switching was an unnoticed and accidental property."

I certainly don't know all "the early literature" but I would be remiss if
I did not point out that there were people who foresaw the problems of
internetwork scalability -- and actively worked on solutions.  In
particular, 15 years before that RFC (in 1981) Yogen Dalal wrote about the
migration from the first generation Xerox Ethernet and internet
architecture, to the second generation of each -- and how the desire to
scale led (at the suggestion of Will Crowther) to the need for a large,
flat address space and associated routing techniques:

"We have described our reasons for choosing absolute host numbers in
internet addresses, and for using them as station
addresses on the Ethernet channel. The host number space should
be large enough to allow the Xerox internet architecture to have a
life span well into the twenty-first century. 48 bits allow for
140,737,488 million physical hosts and mulitcast IDs each. We
chose this size based on marketing projections for computers and
computer-based products, and to permit easy management of the
host number space."
"We expect that the production of microcomputer chips will
increase in the decades that follow, and there will be
microprocessors in typewriters, cars, telephones, kitchen appliances,
games, etc."
     https://dl.acm.org/doi/pdf/10.1145/1013879.802680

This was a really expansive vision about internet scaling and necessary
routing techniques, written about 45 years ago.  Yet Yogen and his team
didn't just think about a "global-scale internet" -- they were already
building one.

John

PS:  While the paper was published in 1981, the actual work had been going
on for some years.  It was unfortunate at the time that we could not share
the detailed work, but in general discussions with the internetworking
community we tried to leave a trail of early breadcrumbs, which an
interested reader might have been able to follow.  See for example, from
IEN 19 in 1978:
"The Implications of Hierarchical vs. Flat Address Spaces
It should be apparent that the structure of the address space is of central
importance: it is the major element of commonality in such an
[internetworking] environment, and can have a profound influence on the
naming mechanism "above" and the routing mechanism "below".
An address space can be partitioned in a hierarchical manner, or left as a
single uniform space. Use of a flat address space implies that:
    1. The address given to any resource must be unique over the whole
    domain; and,
    2. There is no structure to the address which might aid the routing
    process; instead, the routing mechanism must be able to handle all
    addresses, without segmenting them into parts (i.e., there is no
    area code [network number])."