Re: [dev] [ANN] CGD - Ultra-minimalist HTTP and FastCGI wrapper for CGI programs.

From: Kurt Van Dijck <kurt.van.dijck_AT_eia.be>
Date: Wed, 19 Sep 2012 19:36:13 +0200

On Wed, Sep 19, 2012 at 04:43:17PM +0200, Uriel wrote:
> On Wed, Sep 19, 2012 at 4:10 PM, Stephen Paul Weber
> <singpolyma_AT_singpolyma.net> wrote:
> > Somebody claiming to be Uriel wrote:
> >>
> >> CGD runs as a FastCGI wrapper (to be used with nginx or similar web
> >> server) or as a standalone HTTP server, handing over all requests to a
> >> given CGI script.
> >
> >
> > Is this just a wrapper for compatability (which seems odd, since most
> > servers also speak CGI), or does it do prefork stuff to improve performance?
>
> Sadly not all servers speak CGI this days, most notably nginx, and
> others often have broken CGI support.
>
> Also CGD can act as a extremely minimal HTTP server itself to directly
> host "stand-alone" CGI scripts without a 'real' web server.
>
> The server itself uses goroutines which are very efficient, so has no
> need to pre-fork, the called CGI script can't be 'preforked' because
> before you fork a CGI script you have to set the relevant environment
> variables for the request.
>
> Still, forking is never the bottleneck (werc itself performs hundreds
> of fork per request) and I wouldn't be surprised if CGD is not more
> efficient than the CGI-handling code in most web servers, but if your
> web server already supports CGI i wouldn't really bother.

I read somewhere that forking looks like a problem in combination
with multithreaded webservers, since all threads are cloned, in order
to be thrown away after during exec'd.
And so FCGI deamons emerged: push the fork away from the threaded process.

There seemed to be no need to use FCGI for single threaded webservers.

Kurt
Received on Wed Sep 19 2012 - 19:36:13 CEST