Proposed Change to Host/IMP Spec to Eliminate Marking
BBN proposes sending the host/IMP message leader and text as two consecutive messages to end marking problems.
Crowther, W.R., "Proposed Change to Host/IMP Spec to Eliminate Marking", RFC 67, DOI 10.17487/RFC0067, January 1970, <https://www.rfc-editor.org/info/rfc67>.
@misc{rfc67,
series = {Request for Comments},
number = 67,
howpublished = {RFC 67},
publisher = {RFC Editor},
doi = {10.17487/RFC0067},
url = {https://www.rfc-editor.org/info/rfc67},
author = {W.R. Crowther},
title = {{Proposed Change to Host/IMP Spec to Eliminate Marking}},
pagetotal = 1,
year = 1970,
month = jan,
} RFC text
Network Working Group W. Crowther
Request for Comments #67 BBN
[IMPORTANT]
Proposed Change to Host/IMP Spec to Eliminate Marking
Before he left for Norway, Dave Walden observed that there was a
simple change to the Host/IMP specifications which would eliminate all
"marking" problems. The change is to separate the leader from the text
of regular messages, and then to send them as consecutive messages
instead of one packed message. The text would then start on a word
boundary no matter what the word lengths of the machines involved. The
process would take place on both IMP-to-Host and Host-to-IMP messages.
We can see no disadvantages to this scheme, and a sampling of Host
reactions range from neutral to wildly enthusiastic. Therefore, we
tentatively plan to implement this change. We recognize that this
change will involve Host program changes, and that the new release must
be scheduled to coincide with those changes. We propose providing
modified Host spec text along with a one week notice of the new release.
If anyone sees a flaw in this scheme, or is unhappy about it, please
contact BBN quickly, since a new version of the IMP program is already
in the works.
[ This RFC was put into machine readable form for entry ]
[ into the online RFC archives by Kai Henningsen 5/97 ]
AI ELI5
On this page
§1The Marking Problem
W. Crowther of BBN reports that Dave Walden, before leaving for Norway, observed a simple change to the Host/IMP specifications that would eliminate all of the so-called marking problems. The text does not define marking further.
The aim is that message text should start on a word boundary no matter what the word lengths of the machines involved are.
Different computers chop data into different sized pieces, so things did not line up neatly. The idea is to make sure every message starts exactly at the start of a piece.
§2Sending the Leader Separately
The change is to separate the leader from the text of regular messages and send them as consecutive messages instead of one packed message. Because the text then starts on a word boundary, the sizes of the hosts' words stop mattering.
The process would take place on both IMP-to-Host and Host-to-IMP messages.
Instead of writing the address and the letter on one sheet, you put them on two sheets and send them one after the other.
§3Reaction and Rollout
BBN could see no disadvantages, and a sampling of Host reactions ranged from neutral to wildly enthusiastic. It therefore tentatively planned to implement the change.
The note recognizes that the change requires Host program changes, so the new release must be scheduled to coincide with them. BBN proposed providing modified Host spec text along with a one week notice of the new release, and asked anyone who saw a flaw to contact BBN quickly, since a new version of the IMP program was already in the works.
The people who made the network boxes asked, does anybody hate this idea? Nobody did, so they planned to do it and promised to warn everyone a week ahead.