Message-Id: <199702072206.RAA00643@u4-138.openmarket.com> To: Jim Johnstone <firstname.lastname@example.org> Subject: Re: why not wrap fastcgi servers in a shell script? In-Reply-To: Your message of "Wed, 05 Feb 1997 17:11:42 EST." <199702052211.RAA11759@seahawks.lexis-nexis.com> Date: Fri, 07 Feb 1997 17:06:31 -0500 From: Stanley Gambarin <gambarin@OpenMarket.com> > > I start a fastcgi server like this: > AppClass /project/tpf95/data1/dw/usr/jimj/www/fcgi/fcgiEcho.fcg -processes 1 > > where fcgiEcho.fcg is a trivial ksh script to setup the environment: > > #!/bin/ksh > . ./dw.env > exec ./fcgiEcho > > where dw.env is the project standard env file setup script and fcgiEcho is > the fast-cgi server (pretty much straight from the developers kit). > > This all seems to work fine, even for our real servers. Is there some > reason, though, that its a bad idea? > > TIA, Jim Jim, I don't know if that could be a possible problem, but what happens when the fastCGI application runs under one user but the script sets a different euid. It may also turn up some problems with authorization. For the general case, you should not use such setup since it causes an extra fork()/exec() combo, which is not very costly, but can easily be replaced by the PassEnv and SetEnv directives, which are provided by Apache in mod_env. Just a piece of my mind. Stanley.