Mailing List Archive

Consensus on error CGI variable names?
Since 1.4 should be released this week (fingers-crossed), I think we
need a consensus on the variable names. After it comes out, its going
to be much harder to change.

To my knowledge, Apache still uses REDIRECT_, while 1.4 uses ERROR_
The reasons being that the error document has been redirected to by
the request or because the request was in error. Especially with B18,
which means legitimate redirects don't die(), I'm not sure which is more
accurate.

Another thing that maybe should be included is a variable which says which
of the password domain was violated, since this information could be useful
to the script writer as well. Just a thought.

Brandon
--
Brandon Long (N9WUC) "I think, therefore, I am confused." -- RAW
Computer Engineering Run Linux 1.1.xxx It's that Easy.
University of Illinois blong@uiuc.edu http://www.uiuc.edu/ph/www/blong
Don't worry, these aren't even my views.
Re: Consensus on error CGI variable names? [ In reply to ]
> Since 1.4 should be released this week (fingers-crossed), I think we
> need a consensus on the variable names. After it comes out, its going
> to be much harder to change.

> To my knowledge, Apache still uses REDIRECT_, while 1.4 uses ERROR_
> The reasons being that the error document has been redirected to by
> the request or because the request was in error. Especially with B18,
> which means legitimate redirects don't die(), I'm not sure which is more
> accurate.

The Apache philosophy is that all redirects now contain information
about the eariler request, held in the REDIRECT_ prefixed variables.
Custom error/problem responses can be configured to use a local
redirect, and as such, they are no different to any other type of
redirect.

As I've said before, the term "ERROR_" is potentially misleading since
not all calls to die() are a result of an error, e.g. authentication
failure isn't a server error, it's just a denial of service based on
a lack of authentication.

*I* don't want to see Apache use "ERROR_", but I'm prepared to be
overruled by the majority. Only Randy and I expressed any preference
regarding the name, and "REDIRECT_" was chosen. I assumed everyone
else on this list was either happy with the name, or just didn't
care what it was called.

> Another thing that maybe should be included is a variable which says which
> of the password domain was violated, since this information could be useful
> to the script writer as well. Just a thought.

It's proabably worth passing on any info that die() was given about the
problem/error. If I undertand what you're suggesting, then this would
be included.

Maybe "REDIRECT_CAUSE" or "REDIRECT_SYMPTOM" ?

I see no problem with using "REDIRECT_" as the prefix. The web master
will know that (s)he's using the local redirect feature to intercept
problems/errors, so why not refer to the new variables by this name ?..
and if NCSA ever follows Apache and adds CGI variables for other redirects,
you'll need to name them.


robh
Re: Consensus on error CGI variable names? [ In reply to ]
[REDIRECT vs ERROR]
> As I've said before, the term "ERROR_" is potentially misleading since
> not all calls to die() are a result of an error, e.g. authentication
> failure isn't a server error, it's just a denial of service based on
> a lack of authentication.
>
> *I* don't want to see Apache use "ERROR_", but I'm prepared to be
> overruled by the majority. Only Randy and I expressed any preference
> regarding the name, and "REDIRECT_" was chosen. I assumed everyone
> else on this list was either happy with the name, or just didn't
> care what it was called.

REDIRECT is better, for the reason stated above.

Ay. [maintaining a low profile right now 'cuz work is *mental*]

Andrew Wilson URL: http://www.cm.cf.ac.uk/User/Andrew.Wilson/
Elsevier Science, Oxford Office: +44 01865 843155 Mobile: +44 0589 616144
Re: Consensus on error CGI variable names? [ In reply to ]
On Mon, 1 May 1995, Andrew Wilson wrote:
> [REDIRECT vs ERROR]
> > As I've said before, the term "ERROR_" is potentially misleading since
> > not all calls to die() are a result of an error, e.g. authentication
> > failure isn't a server error, it's just a denial of service based on
> > a lack of authentication.
> >
> > *I* don't want to see Apache use "ERROR_", but I'm prepared to be
> > overruled by the majority. Only Randy and I expressed any preference
> > regarding the name, and "REDIRECT_" was chosen. I assumed everyone
> > else on this list was either happy with the name, or just didn't
> > care what it was called.
>
> REDIRECT is better, for the reason stated above.
>
> Ay. [maintaining a low profile right now 'cuz work is *mental*]

I'll throw in my +1 as well for REDIRECT. Consider, NCSA, that with
HTTP/1.1 an imagemap script could redirect internally to a CGI-script
rather than force the client to make a second request, and in that
situation the redirect is not an error at all (and that the script might
need the info that the imagemap script had access to).

Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com brian@hyperreal.com http://www.[hyperreal,organic].com/