why export shell variables

[utzoo!decvax!duke!chico!harpo!jerry]

I am putting this in net.v7bugs because the item I am replying to
appeared there.  However, the requirement that variables be explicity
exported before seting them changes the environment is not a bug.
It was, I have been told, a deliberate decision and there are reasons
for it.

The bourne shell is dealing with environment values, and its own local
variables.  Unlike csh, it uses the same syntax to set these different
kinds of values.  Therefore, it requires a declaration of some sort
to distinguish them.  It uses "export".  I suppose requiring local
variables to be declared would also be reasonable.  Local variables
are important in writing shell procedures and it would be a real loss
to eliminate the distinction between them and environment variables.

Problems arise because people frequently set up an interactive shell
(e.g. via an escape from the editor) and don't realize that they must
export a variable before they can change its value in the environment.
For this reason, I would prefer to make the distinction in
the assignment syntax (ala csh).  But, I consider the current situation
preferable to one in which a shell procedure might inadvertantly
interfer with the transfer of a value from outside (i.e.
the environment) to a command that it invokes.

						Jerry Schwarz
						harpo!jerry