httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brian Behlendorf <br...@hyperreal.com>
Subject Re: FastCGI and SSL. (fwd)
Date Thu, 22 Aug 1996 21:24:21 GMT

I'll put the fastcgi.c in /apache/incoming (it was 134K)

	Brian


---------- Forwarded message ----------
Date: Tue, 20 Aug 1996 16:11:58 +0100
From: Nigel Metheringham <Nigel.Metheringham@theplanet.net>
To: Mark Brown <mbrown@OpenMarket.com>
Cc: fastcgi-developers@OpenMarket.com, new-httpd@hyperreal.com,
    ssl-users@mincom.oz.au
Subject: Re: FastCGI and SSL. 

} 
} Michael Smith asks:
} 
}     Has anybody else out there tried to use fastCGI and SSL in apache?
} 
} Yes, Nigel Metheringham has tried and succeeded; see
} 
}     http://www.fastcgi.com/mail-archive/0096.html

Fame at last :-)

} Ben Laurie is correct that mod_fastcgi violates the
} buff abstraction, which breaks SSL; Nigel's message includes
} a patch.

However I am told people had trouble with that patch *and* my work 
was done on apache 1.1b4 using a pre-Ben version of SSL -- I got SSL 
working and passed the patches to Ben.  I have not looked at what Ben 
released, although I know he at least started from my material... 
having said that SSL did not impact mod_fastcgi.c at all, it was just 
the buff/direct-fd usage that needed cleaning up.

I am enclosing the most recent hacked version of mod_fascgi.c I have. 
 I am told that it doesn't work right on Apache 1.1.1 with SSL - the 
error described makes it sound as though the bread/bwrite functions 
within buff.c have had a change of name, in which case the fix is 
easy.  It also does not attempt to support pre-1.1 apache - feel free 
to add support back in.  I am quite willing to let this go out under 
GPL or whatever, just remember me if you make *lots* of money off it 
:-) .  I am really overloaded right now so cannot put any work into 
making things better :-(

There may still be one bug I know of left, although I think I nailed 
it finally.... under very rare circumstances it used to get itself 
into some deadly embrace where it didn't think it had all the data, 
yet the other end was waiting for data to be sent.

} I would like to incorporate the essence of Nigel's patch
} into the released version of mod_fastcgi.  The only downside
} is that doing so prevents a FastCGI application from
} engaging in filter-like behavior: writing large amounts
} of data to stdout before fully consuming stdin.

Not sure if thats true actually... it should be able to handle input 
and output independantly unless I broke the select statement :-)

	Nigel.



Mime
View raw message