When did people start writing Readme files?
It seems that pretty much all programs have this file, regardless of the format.
Is there any documented first use of this document?
I don't know of a canonical first use. The Jargon File describes the README as:
Hacker's-eye introduction traditionally included in the top-level directory of a Unix source distribution
So, advances on July 1977 are welcome!
I took an Operating Systems class this last year and remember my professor telling me that they had README's (actual physical printed files) for all of their punch cards and mag tape and pretty much anything else that was a "program". At that time you really needed one because of the labourous process that was involved with getting the created, ran, and everything else. These README's sometimes also included the actual printouts of how the punch cards were supposed to be punched as a form of error checking and debugging.
The convention apparently also follows the old system in that with all the punch cards a "reem" of paper was attached with the statement README in caps printed on it, this had all of the instructions for use and loading of the punch cards into the system.
For a time reference, this would have been in the 60's. But the punched card system has been around since the 1700's for controlling of "automated" looms, so maybe even as old as that.
Found via this alt.folklore.computers thread:
I'm betting some document accompanying one of the PDP-8 operating systems or software packages will probably be the oldest.
The oldest I found in the PDP-10 archives is the UCI LISP "READ.ME" from the 4th DECUS library tape, with a timestamp of 27-Mar-1975:
Unfortunately, many of the earlier mini OS's do not support a [unique] timestamp in the tape and/or disk directory structure, so it's hard to tell when a document was really last edited.
Not only does the file have a 27 March 1975 timestamp, but it's handily signed and dated in the file itself:
UCI LISP Random Notes [...] Should problems arise with this system, please communicate them to the Department of Information and Computer Science at UCI; they will be directed to whomever is currently managing the UCI LISP system. William J. Earl Department of Information and Computer Science University of California Irvine, California 27 March 1975
[README.TXT is the DOC file for SPICE/SINC/SLIC]
This failsafe tape contains the circuit analysis programs:
SPICE SINC and SLIC
described in the Applications Software Bulletin Volume 4.
SPICE requires FORTRAN-10 version 4 because of its use of Right adjusted Holerith data. Executes in about 47K.
it also includes this file, the FOROTS to go with the SAVes and the source for SECOND.MAC, the timing routine. SPICE is broken into three parts: 1SPICE.FOR, 2 and 3.
There is a printed document to describe each of the programs. These are included in the DECUS packet. The documentation and programs were origionally developed by the E.E. department of the Univ. of Calif. at Berkley on a CDC 6400. Except to convert the FORTRAN to the DECsystem-10 no changes have been made to the programs. For the test data SLIC and SINC have shown a slight variation with respect to the 6400, SPICE shows no variation.
Good luck! Ashley Grayson 27-NOV-74
[end of README.TXT]