Re: CGI???

Chris Adie (
Wed, 22 Nov 1995 10:19:58 +0000

David Robinson <> wrote:
> Chris Adie <> wrote:
> > ...
> > So, avoid
> > implementation-specific terms like "environment variables", and call them
> > (say) "CGI variables" instead.
> I disagree. a CGI `specification' wouldn't be much of a specification if it
> didn't allow a programmer to write a working program based on it.

That is arguable - splitting a specification into system-independant and
system-dependant parts is not uncommon.

> From a practical standpoint, a separate platform-specific spec seems pretty
> pointless when, for example, the Unix-specific spec is only 9 lines of text.

That is because CGI was developed for UNIX systems. I'm not going to fight
hard for the UNIX-specific spec to be physically in a separate document;
however I do believe the CGI spec should define the semantics of CGI variables
and CGI data i/o in a non-implementation-specific way.

In fact, your spec very nearly achieves this. When talking about data
input/output from the CGI script, you say "Unless defined otherwise, this
will be via the `standard input' file descriptor." Why not use the same
form of words for CGI variables?


Chris Adie Phone: +44 131 650 6773
Edinburgh University Computing Service Fax: +44 131 650 6552
James Clark Maxwell Building Email:
Kings Buildings
Edinburgh EH9 3JZ, United Kingdom