nph- deprecation

[sa...@c2.net]

	Because of the dangers of nph-, as described here earlier,
I proposed:

1) an option to disable nph: NphDisable
2) making my unbuffered cgi patch a "supported patch"

	Should I work on this? I'd like to let people avoid the nph-
security problem, and deprecating nph- has always been one of my
goals.

	I do think #1 is worth going into the beta, because it is a
security thing. #2 is too iffy/featurish to put in right now.

Re: nph- deprecation

[ra...@lerdorf.on.ca]

> > 2) making my unbuffered cgi patch a "supported patch"
> 
> +1.
> 
>      Brian

If we are adding that, how about an rflush() call so I can stop messing
with the BUFF stuff from my modules?  Personally I see little need
for unbuffered output if it were easy to flush the buffer.

-Rasmus

Re: nph- deprecation

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

On Wed, 8 Jan 1997 sameer@c2.net wrote:
> 1) an option to disable nph: NphDisable
> 	I do think #1 is worth going into the beta, because it is a
> security thing. 

Hmm... what about instead, closing the fd to stdin to the nph script after the
script starts outputting data?  That way the nph script can't answer a second
kept-alive request and start spoofing things.  Will that work?  I'd +1 that as
a security fix for 1.2.  Disabling NPH scripts isn't enough for 1.2 imho.

> 2) making my unbuffered cgi patch a "supported patch"

+1.

	Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com  www.apache.org  hyperreal.com  http://www.organic.com/JOBS