Hey Rob, welcome back!
Rob McCool says:
>
> Please. Magic directory names and magic extensions are the domain of the
> server; it's up to the server authors to decide how they're going to
> distinguish between documents and scripts. I happen to view our decision to
> make the distinction between script execution and document retrieval
> transparent as a feature, ...
I agree completely that distinguishing documents from scripts should
be up to the server author. My main gripe (as I have reitered many
times) is that I feel the protocol is forcing me to use an implementation
(parsing via stat() ) that I don't like when a different protocol
design wouldn't limit my choices this way. How you implement your
server is none of my business. I am not that familiar with your server
and I never intended to be critical of your implementation.
Anyway, as I said above, I give up. I understand and accept the
argument that it is now too late to make changes in the protocol.
As for the other arguments I am not so much persuaded as inundated.
As I understand what Sanders and Thau are saying, the syntax of the
path part of a URL should be be deliberately ambiguous as to whether
the object is a directory file or script and and also ambiguous as to
what part represents a path and what part represents arguments. This
is done intentionally so the maintainer can change his or her mind
about all these things at a later date.
It would be helpful if the CGI spec explained that this ambiguity is
intentional, what the rationale for it is, and that the designers
believe that this flexibility is worth the cost of disambiguating
the URL at the server. A paragraph explaining this would have saved
me a lot of time.
On the other hand, it has been fun watching Rob blow a whole day
working his way through this discussion :)
John