a problem with terminology Thomas Bushnell BSG (12 Jan 2010 04:24 UTC)
|
Re: a problem with terminology
Derick Eddington
(12 Jan 2010 05:46 UTC)
|
Re: a problem with terminology
Thomas Bushnell BSG
(12 Jan 2010 06:08 UTC)
|
Re: a problem with terminology
Derick Eddington
(12 Jan 2010 06:36 UTC)
|
Re: a problem with terminology
Derick Eddington
(12 Jan 2010 07:12 UTC)
|
a problem with terminology Thomas Bushnell BSG 12 Jan 2010 04:05 UTC
This is a more minor concern than my previous note, but also important. The srfi uses the term "path" in mutual confusing ways. Richard Stallman impressed upon me--and I think he's quite right--that Unix has traditionally used the word "path" in two quite different senses. Sense one means by a path a sequence of directory names, separated on Unix with a slash. When we say that "/usr/include/stdio.h" is a path, this is sense 1. Sense two means a list of sense-one paths, often written with a colon as separator between them, identifing a list of directories to search to find a file. When we say that your PATH is /bin:/usr/bin:/usr/local/bin, this is a sense-2 path. The GNU Project has abandoned this systematic ambiguity and never uses the word "path" in sense-1, sense there is the perfectly good "filename" already. I believe that Posix as well never uses "path" in sense-1, but I can't recall exactly. (Historically, "filename" referred in Unix to the entry in a specific directory, but those days are gone.) The srfi introduces the concept of a "library file path", which is a sense-1 path, and which is essentially a specification of the format of an r5rs/r6rs "filename", and a re-specification of a Unix filename. Can I request that the srfi drop the attempt to call filenames "path names", and stick to "path" specifically in the context of a search path? Thomas