Re: server cgi include bug (fwd)

[Rob Hartill <ha...@hyperreal.com>]

Forwarded message:
> From sharp@synasoft.com  Thu Aug  3 19:36:35 1995
> Message-Id: <19...@synasoft.interaccess.com>
> Subject: Re: server cgi include bug
> To: apache-bugs@hyperreal.com
> Date: Thu, 3 Aug 1995 21:40:15 -0500 (CDT)
> In-Reply-To: <19...@taz.hyperreal.com> from "Rob Hartill" at Aug 3, 95 09:20:33 am
> From: Daryn Sharp <sh...@synasoft.com>
> Reply-To: sharp@synasoft.com
> X-Disclaimer: I speak solely for myself...
> X-Mailer: ELM [version 2.4 PL24]
> Content-Type: text
> Content-Length: 1024      
> 
> >CGI scripts *must* have headers. They're not CGI if they don't
> >produce a header.
> >If you want to include output of a script
> >that doesn't produce headers, you should use exec cmd=""
> 
> This is not how NCSA's server works, that's why I call it a bug.  The problem
> with using "exec cmd" is that you must know the absolute path to the script,
> but "exec cgi" allows you to use a virtual path.  I can't afford to use
> absolute paths since the scripts must be portable with no modifications.
> 
> With NCSA's server, I wrote docs that had "exec cgi" includes imbedded
> in the middle of a document.  Now they don't work with Apache, which
> advertises itself as a "drop-in replacement".  This is not so, and
> therefore it's a misleading statement.
> 
> Either Apache needs to conform to NCSA standards, or a mechanism to
> exec cmds via virtual paths needs to be added.  The former is highly
> preferable for compatiblity reasons.  a "drop-in replacement" should
> be exactly that.
> 
> --- Daryn Sharp  <sh...@synasoft.com>  "I speak solely for myself"
> 

I can live with that inconsistency.

The virtual path thing might be worth considering.

Any thoughts ?


rob

Re: server cgi include bug (fwd)

[Brian Behlendorf <br...@organic.com>]

On Fri, 4 Aug 1995, Rob Hartill forwarded:
> > >CGI scripts *must* have headers. They're not CGI if they don't
> > >produce a header.
> > >If you want to include output of a script
> > >that doesn't produce headers, you should use exec cmd=""
> > 
> > This is not how NCSA's server works, that's why I call it a bug.  The problem
> > with using "exec cmd" is that you must know the absolute path to the script,
> > but "exec cgi" allows you to use a virtual path.  I can't afford to use
> > absolute paths since the scripts must be portable with no modifications.

So, the solution is obviously to allow "exec cmd" to use relative paths.  
That's how #include works too.

> > With NCSA's server, I wrote docs that had "exec cgi" includes imbedded
> > in the middle of a document.  Now they don't work with Apache, which
> > advertises itself as a "drop-in replacement".  This is not so, and
> > therefore it's a misleading statement.

Hahaha.

> > Either Apache needs to conform to NCSA standards, 
[er, bugs]

> > or a mechanism to
> > exec cmds via virtual paths needs to be added.  

This is what I'd prefer

> > The former is highly
> > preferable for compatiblity reasons.  a "drop-in replacement" should
> > be exactly that.

"...highly preferable because I'm lazy and don't like fixing my bugs."

Sorry, busy Friday.  The docs are still in progress.  RobT, you can 
announce anytime, it'll be a week before the documentation is fully 
complete but I'll work on it this weekend.  I will have 0.8.7 on 
hyperreal and organic before the end of the day.

	Brian

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