"But my NCP costs $500 a day"
A standard host status service to reduce network polling overhead
Bressler, R.D., ""But my NCP costs $500 a day"", RFC 425, DOI 10.17487/RFC0425, December 1972, <https://www.rfc-editor.org/info/rfc425>.
@misc{rfc425,
series = {Request for Comments},
number = 425,
howpublished = {RFC 425},
publisher = {RFC Editor},
doi = {10.17487/RFC0425},
url = {https://www.rfc-editor.org/info/rfc425},
author = {R.D. Bressler},
title = {{"But my NCP costs $500 a day"}},
pagetotal = 1,
year = 1972,
month = dec,
} RFC text
Network Working Group Bob Bressler (BBN)
Request for Comments: 425 Dec 1972
NIC #13010
"But my NCP costs $500 a day..."
Now that networking is becoming commonplace, people are beginning to measure
the cost of their network software and network use. A not insignificant
portion of this use turns out to be random prodding and poking of one system
by another. An example of this is the "Survey" mechanism that many sites are
constructing to see "who's up." Several systems have noted that this one
mechanism alone accounts for a log of their overhead, and is not chargeable to
any given user.
In response to this particular problem, MIT-DMCG and UCLA-NMC, who have both
been doing surveys for some time, have agreed upon a standard format for host
survey data and will make it available to the network community. They will be
joined by a third site, and will provide a standard socket to obtain this
information. The data will be a concise character stream so as to be both
human and machine readable.
The exact format of the data and the socket number will be forthcoming.
The intent of all this is for all other sites to stop surveying on their own
and get this data from one of the participating "Official Surveyors" either
prompted by a user's request or on a periodic basis, whatever they wish.
This should, itself, help considerably to eliminate a lot of the network
'noise factor.' But, in addition, we should all try to be more aware of the
cost of this type of traffic and try to avoid repeated interactions with a
given Host without some expressed consent - be it an account number of some
specific experiment.
BB/ph
AI ELI5
On this page
§1The Cost of Host Surveys
As network use became more common, sites began measuring the cost of their network software and network usage. They discovered that a significant amount of network overhead came from host surveys—mechanisms that many sites built independently to check 'who's up' on the network. Several systems found that this survey activity alone accounted for a large portion of their overhead and could not be charged to any specific user.
Every computer on the network kept asking other computers if they were awake, wasting a lot of time and effort.
§2Standard Host Survey Data
In response to this problem, MIT-DMCG and UCLA-NMC, both of which had been doing host surveys for some time, agreed upon a standard format for host survey data and made it available to the network community at a standard socket number. A third site joined this effort, forming a group of Official Surveyors that would perform the surveys on behalf of the entire network.
Three big computer sites agreed to ask if other computers were awake, and everyone else could just ask them instead of asking directly.
§3Reducing Network Noise
The goal was for all other sites to stop surveying hosts independently and instead get host status data from one of the Official Surveyors, either prompted by a user's request or on a periodic basis. The standard format for the data was designed to be concise and both human-readable and machine-readable, making it efficient to transmit and process. This approach would help eliminate much of the network traffic that contributed to the 'noise factor' on ARPANET.
By having just a few central check points instead of everyone asking, the whole network uses less effort.
§4Network Cost Awareness
Beyond the standardized survey service, the RFC also urged all sites to become more aware of the cost of network traffic and to avoid making repeated interactions with a given host without some expressed consent, such as an account number or some specific experiment justification. This awareness of cost and efficiency became important as network use grew.
Everyone should think about how much effort they are using on the network and ask permission before using someone else's computer a lot.