How big is the Internet?

Sean Donelan sean at
Fri Aug 16 19:21:07 UTC 2013

On Fri, 16 Aug 2013, bmanning at wrote:
> On Fri, Aug 16, 2013 at 12:37:20AM -0400, Sean Donelan wrote:
>> Even the researchers at the Library of Congress, if you give them
>> enough beer and beg them enough, will eventually give you an estimate
>> about the Library collection size as of the end of the last year.
>> What so special about the Internet that it can't be measured?
> The problem is that is can be measured, along a large number variables.
> The LOC question, How Big?  Might be linear shelf space, sqft, number of
> items, number of warehouses, number of employees, budget, etc.  The
> base question, How Big needs a qualifier or two.

So, in the context of the LOC question about using it as a unit of 
measurement for comparison with storage size or transmission volume; which
of those things are information that can be transmitted or electronically

If I asked "how big is an elephant?" some zoologists would look for 
ways the question can't be answered like "elephants grow from birth to 
death, have different species, may have illnesses, have not measured every
elephant, etc."; others might give an answer like "Adult male elephants 
usually stand ten to thirteen feet tall and can weigh seven to twenty-six 
thousand pounds. Females elephants tend to be smaller smaller."

> Same with the Internet.  How big makes no sense.  How much traffic begs
> the question of measured from where.  A unique attribute of IP based
> transport is that -as far as I know- there is no measurement point between
> -every- pair of nodes that might exchange traffic.

That is true of most transportation networks: roads do not have 
measurement points between every destination point, oceans do not have a
measurement point between every port.  Intangiable things like the 
"economy" don't have measurement points at every economic transaction.

Yet there are relatively accepted estimates of the size Gross Domestic 
Product, annual miles driven on US Highways, shipping between countries.

> And since the instrumentation does not exist, you'll never get the numbers.
> Select other vectors and the problem remains, the instrumentation is poor
> or non-existant.
> Any numbers that are derived are incomplete and/or estimates.
> Pick your poision.

Ed Felten gave the keynote talk at Usenix Security this week. One
of the examples he gave was a out-of-town friend asking a technologist
for recommendations for a good resturant.  Hilarity ensured.

If you wonder why other people don't ask technologists for answers,
Dr. Felten's talk is a good starting point.

More information about the NANOG mailing list