Tony Abou-Assaleh <[EMAIL PROTECTED]> writes: > In principle, I am totally against intentionally not documenting features > of a program that exist for a purpose other than debugging the program.
The documentation is the contract between the program and its users. In some cases, it is wise to not over-specify that contract, to give the implementation some freedom. There are many, many places where the grep manual does not document accidental features of grep. This is not necessarily a mistake, any more that it is a mistake that the C standard intentionally does not specify the exact value of the maximum 'int' value, or the exact order of evaluation of an expression. If someone wanted to take the time to document all these accidental features of 'grep', in a section clearly labeled "EVERY STATEMENT IN THIS SECTION IS TRUE ONLY BY ACCIDENT, AND IS SUBJECT TO CHANGE IN THE VERY NEXT GREP RELEASE; USE AT YOUR OWN RISK", then it might be useful to do that. But that someone would have to commit to keeping the documentation of accidental features up to date. As far as I know, nobody is seriously prepared to do that.
