[ih] Bandwidth v Capacity [Early Internet Report when Vint was at Stanford (and DARPA PI)]
Bill Woodcock
woody at pch.net
Sun May 31 06:55:49 PDT 2026
Let me preface this by saying that I don’t imagine that everyone who uses these terms uses them uniformly or conscientiously. The distinction _I_ would make between “bandwidth” in the business sense, and “capacity” is that bandwidth is actually used, whereas capacity is a latent capability of use. i.e. a broadband customer may have an Ethernet connection with 1gbps of capacity, but the “amount of bandwidth” is the portion of that capacity which they actually exercise. Thus ISPs have an “oversubscription ratio” between their downstream sold capacity and their upstream transit + peering capacity, but the sum of their downstream bandwidth sold is approximately equal to the sum of their upstream transit + peering, albeit the bandwidth varies (typically diurnally, weekday/weekend, and August-in-Europe). Whereas the capacity is fixed at any given time, subject to provisioning, hardware constraints, ditches getting dug, poles getting planted, etc.
I try to use these terms consistently within my own writing and thinking and conversation, and there are many other people with whom I have a shared understanding of their meanings; but I also recognize that there are a lot of people who use both terms naively, or who use them with other meanings in mind. I don’t know whether those people can have meaningful conversations with anybody about these topics, though. :-)
> On May 31, 2026, at 15:36, John Day via Internet-history <internet-history at elists.isoc.org> wrote:
>
> You are undoubtedly right and it is to some degree just an annoyance.
>
> The other misconception that goes with this is that if the ISP advertises X ‘bandwidth/capacity’ should the customer expect to have X available to them all the time? How many other customers could the ISP support if that were the case?
>
>> On May 31, 2026, at 08:59, Bill Woodcock <woody at pch.net> wrote:
>>
>> I do understand the difference between bandwidth and capacity and throughput and goodput and so forth, as used as terms-of-art within our industry. But I also recognize that “bandwidth” is also a term-of-art in the _business_ of the Internet, meaning “the thing that ISPs sell.” It doesn’t make sense to complain that they need to change or use a different word or use a whole paragraph to describe it. And we can’t really expect that usage not to leak back into the rest of the industry, when it’s a more commonplace usage than the “range of frequency” term-of-art.
>>
>> So I sympathize, but I don’t think this is a hill worth dying on.
>>
>> -Bill
>>
>>
>>
>>
>>> On May 31, 2026, at 12:48, John Day via Internet-history <internet-history at elists.isoc.org> wrote:
>>>
>>> Brian,
>>> I totally agree with you. In fact, I have been known to use a question like that on exam. We are sloppy in our use of language.
>>>
>>> Besides bandwidth vs capacity, I have found that there are professionals out there who believe that analog and digital signals are fundamentally different. Not that a digital signal is one that hasn’t traveled far enough to betray that it is really analog underneath.
>>>
>>> And as long as we are cleaning up language, a network graph is not a topology. One could refer to a set of network graphs with a common invariant as a topology, but a single network graph is just a graph.
>>>
>>> Take care,
>>> John
>>>
>>>> On May 31, 2026, at 06:32, Brian Carpenter via Internet-history <internet-history at elists.isoc.org> wrote:
>>>>
>>>> I'm sorry if it came across as criticism; I genuinely wonder when the usage
>>>> originated.
>>>>
>>>> (via tiny screen & keyboard)
>>>> Regards,
>>>> Brian Carpenter
>>>>
>>>> On Sun, 31 May 2026, 21:30 vinton cerf, <vgcerf at gmail.com> wrote:
>>>>
>>>>> Rant received; thanks for the tutorial.
>>>>> v
>>>>>
>>>>>
>>>>> On Sat, May 30, 2026 at 11:55 PM Brian E Carpenter via Internet-history <
>>>>> internet-history at elists.isoc.org> wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> Yes, that's a very nice slice of history. It leads me to one of my hobby
>>>>>> horses: when did the solecism of using "bandwidth" to mean "capacity" first
>>>>>> arise? This is something that should annoy every physicist, or anyone who
>>>>>> has read Shannon's foundational paper [1]. Unfortunately, it's become
>>>>>> firmly established in the Internet community and beyond. And it matters
>>>>>> because it quite often creates confusion, particularly in media reports.
>>>>>>
>>>>>> * Bandwidth is measured in herz (cycles per second) and is the frequency
>>>>>> range that a communication channel can transmit.
>>>>>>
>>>>>> * Capacity is measured in bits per second and is the amount of binary
>>>>>> information that a communication channel can transmit.
>>>>>>
>>>>>> There is no fixed relationship between the two (which is more or less the
>>>>>> main point of Shannon's paper). Anyone who ever heard the startup screeches
>>>>>> of a modem should know this.
>>>>>>
>>>>>> To illustrate the issue, page 1 of Vint's report says:
>>>>>>
>>>>>> "During the months of December (1975) and January (1976), we undertook
>>>>>> extensive and detailed timing measurements of the ELF VDH behavior to
>>>>>> ascertain the degree to which VDH performance affected total TCP
>>>>>> bandwidth and delay."
>>>>>>
>>>>>> That (and the other 14 occurrences of "bandwidth") should be "capacity."
>>>>>>
>>>>>> Page 8 says:
>>>>>>
>>>>>> "The actual
>>>>>> line utilization is about 20% in each direction, assuming a nominal 50
>>>>>> kbits/second available full-duplex capacity between ELF and the IMP."
>>>>>>
>>>>>> That's correct usage.
>>>>>>
>>>>>> My favourite sentence is on page 45:
>>>>>>
>>>>>> "Many of the experiments have been
>>>>>> frustrating, owing to a bug of some kind in UCL's buffer allocation scheme
>>>>>> causing them to crash irrevocably when attempting to achieve high
>>>>>> bandwidth."
>>>>>>
>>>>>> If only they had tried to achieve high capacity!
>>>>>>
>>>>>> So, my question is: when did this inaccurate use of "bandwidth" to mean
>>>>>> "capacity" first arise? It was clearly well established by 1975.
>>>>>>
>>>>>> I looked in Donald Davies's book [2], and it only uses "capacity" (I
>>>>>> would expect no less of him). Baran in 1964 [3] used "capacity" correctly
>>>>>> many times. He also used "bandwidth" correctly once [5] and debatably a
>>>>>> second time [6].
>>>>>>
>>>>>> Pierce [4] in 1961 was completely clear on the difference between
>>>>>> bandwidth and capacity (and he learned directly from Shannon).
>>>>>>
>>>>>> Regards/Ngā mihi
>>>>>> Brian Carpenter
>>>>>>
>>>>>> [1] C. E. Shannon, "Communication in the Presence of Noise," Proceedings
>>>>>> of the IRE, vol. 37, no. 1, pp. 10-21, Jan. 1949
>>>>>> [2] D. W. Davies, D. L. A. Barber, W. L. Price & C. M. Solomonides,
>>>>>> "Computer Networks and their Protocols," Wiley, 1979
>>>>>> [3] P. Baran, "On Distributed Communications Networks," IEEE Transactions
>>>>>> on Communication Systems, Vol. 12 No. 1, 1964, pp. 1-9
>>>>>> [4] J. R. Pierce, "Symbols, Signals and Noise," Harper, 1961
>>>>>>
>>>>>> [5] "In a conventional circuit-switched system each of the
>>>>>> tandem links requires matched transmission bandwidths.
>>>>>> In order to make fullest use of a digital link, the post-
>>>>>> error-removal data rate would have to vary, as it is a
>>>>>> function of noise level."
>>>>>>
>>>>>> [6] "Most importantly, standardized data blocks permit
>>>>>> many simultaneous users, each with widely different band-
>>>>>> width requirements to economically share a broad-band
>>>>>> network made up of varied data rate links."
>>>>>>
>>>>>> On 30-May-26 15:13, Barbara Denny via Internet-history wrote:
>>>>>>> While poking around on DTIC, I found this interesting early report on
>>>>>> the Internet while Vint was still at Stanford and Jon Postel was still at
>>>>>> SRI (How many of you knew that? :-))The period of performance is Nov
>>>>>> 15,1975 to Feb 15 1976.
>>>>>>> https://apps.dtic.mil/sti/tr/pdf/ADA024823.pdf
>>>>>>> You can find a little bit more info by using the contract number in
>>>>>> dtic.
>>>>>>> Happy Reading,barbara
>>>>>> --
>>>>>> Internet-history mailing list
>>>>>> Internet-history at elists.isoc.org
>>>>>> https://elists.isoc.org/mailman/listinfo/internet-history
>>>>>> -
>>>>>> Unsubscribe:
>>>>>> https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
>>>>>>
>>>>>
>>>> --
>>>> Internet-history mailing list
>>>> Internet-history at elists.isoc.org
>>>> https://elists.isoc.org/mailman/listinfo/internet-history
>>>> -
>>>> Unsubscribe: https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
>>>
>>> --
>>> Internet-history mailing list
>>> Internet-history at elists.isoc.org
>>> https://elists.isoc.org/mailman/listinfo/internet-history
>>> -
>>> Unsubscribe: https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
>>
>>
>
> --
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history
> -
> Unsubscribe: https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
-Bill
Please consider the environment before using AI to process this email.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: Message signed with OpenPGP
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20260531/521e30b9/attachment.asc>
More information about the Internet-history
mailing list