Re: mod_gzip and static fastcgi applications

Thank you for the advise.

Unfortunately that wasn't what I was looking for.

It may not be practical to just have a fastcgi-only server in all
real-life situations. Whether my configuration is ideal is not the
issue. Clearly there is a problem in mod_fastcgi which is overwriting a
value set by another component. The issue is how to fix this bug and
whether my fix was appropriate.

I'd appreciate if anyone could comment on my question.

Thanks for your time,
/prakash

On Tue, Nov 09, 2004 at 01:41:59PM -0500, Jesus Altuve wrote:
> Hi,
> 
> I suggest you don't compile too much modules with your apache if what you
> want is performance. My advise is to compile a FCGI only server, and use a
> reverse proxy in front with mod_gzip, that way slow clients won't affect you
> and your FCGI environment will be rock stable.
> 
> ----- Original Message ----- 
> From: "Prakash Kailasa" <[email protected]>
> To: <fastcgi-developers-xGejAJT2w6yOE5Ap4OspEtHuzzzSOjJt@public.gmane.org>
> Sent: Tuesday, November 09, 2004 11:15 AM
> Subject: Re: [FASTCGI] mod_gzip and static fastcgi applications
> 
> 
> > Hi,
> >
> > I am sure this list has more important things to talk about, but can I
> > get some response to my question? Please?
> >
> > I just want to know if the modification made to accommodate mod_gzip
> > was appropriate and does not break things otherwise. We have been
> > testing it without any problems so far, and admittedly this is very
> > tiny change, but I don't know fastcgi code well, so I'd feel much
> > better if it is approved by the fastcgi developers.
> >
> > Thanks for your time,
> > /prakash
> >
> > On Thu, Oct 28, 2004 at 12:24:29PM -0400, Prakash Kailasa wrote:
> > > Hi,
> > >
> > > We use fastcgi (at my workplace) for a several perl applications, both
> > > static and dynamic. Recently, we added mod_gzip to the mix.
> > >
> > > The compression works fine in case of dynamically started fastcgi
> > > applications, but the output of static fastcgi apps does not go thru
> > > the compression stage.
> > >
> > > After much debugging, I've found that the fixups function in
> > > mod_fastcgi.c is setting the value of the handler to
> > > FASTCGI_HANDLER_NAME ("fastcgi-script") in case of static
> > > applications, replacing the value set by mod_gzip.
> > >
> > > Here's the relevant part of the code:
> > >
> > >     if (fcgi_util_fs_get_by_id(r->filename, uid, gid))
> > >     {
> > >         r->handler = FASTCGI_HANDLER_NAME;
> > >         return OK;
> > >     }
> > >
> > > The test above fails for dynamic apps, so there would be no change in
> > > the handler (which happened in our case to be mod_gzip_handler) and
> > > everything was fine.
> > >
> > > I looked at the CHANGES file and found this item (under 2.4.0):
> > >
> > >  *) Eliminate the need for SetHandler or AddHandler with static or
> > >     external applications.
> > >
> > > Not knowing the fastcgi code throughly, I am not sure if this item
> > > corresponds to the above code segment. But I am guessing that not
> > > having either of these directives would leave the handler unset and
> > > the fixups function is fixing this situation.
> > >
> > > If I change the above test to include an additional check so that the
> > > handler is fixed only if it is truly NULL,
> > >
> > >     if (r->handler == NULL && fcgi_util_fs_get_by_id(r->filename, uid,
> gid))
> > >
> > > gzip compression works in case of static applications as well.
> > >
> > > My questions: Is this an appropriate fix? Does this break any other
> > > situations? If so, what would be the right fix?
> > >
> > > My environment:
> > >     Apache 1.3.28
> > >     mod_fastcgi 2.4.0
> > >     mod_gzip 1.3.26.1a
> > >     Linux 2.4.25
> > >
> > > I am attaching the patch, if that is of any help.
> > >
> > > Thanks for your time,
> > > /prakash
> > >
> > > -- 
> > > |All language designers are arrogant. Goes with the territory...       |
> > > |        -- Larry Wall                                                 |
> >
> > > --- mod_fastcgi.c.orig 2002-12-22 22:19:14.000000000 -0500
> > > +++ mod_fastcgi.c 2004-10-04 16:36:30.000000000 -0400
> > > @@ -2812,7 +2812,7 @@
> > >
> > >      get_request_identity(r, &uid, &gid);
> > >
> > > -    if (fcgi_util_fs_get_by_id(r->filename, uid, gid))
> > > +    if (r->handler == NULL && fcgi_util_fs_get_by_id(r->filename, uid,
> gid))
> > >      {
> > >          r->handler = FASTCGI_HANDLER_NAME;
> > >          return OK;
> >
> > > ___________________________________
> > > fastcgi-developers mailing list
> > > http://fastcgi.com/fastcgi-developers/
> >
> >
> > -- 
> > |Mollison's Bureaucracy Hypothesis:                                    |
> > |        If an idea can survive a bureaucratic review and be           |
> > |        implemented it wasn't worth doing.                            |
> >
> 
> 
> ----------------------------------------------------------------------------
> ----
> 
> 
> ___________________________________
> fastcgi-developers mailing list
> http://fastcgi.com/fastcgi-developers/

-- 
|You can't have everything.  Where would you put it?                   |
|                -- Steven Wright                                      |

___________________________________
fastcgi-developers mailing list
http://fastcgi.com/fastcgi-developers/