httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rob Hartill <>
Subject 1.2b1 status
Date Thu, 21 Nov 1996 17:39:33 GMT

	Sameer has offered a patch to do unbuffered CGI.
	Ben has said he supports the feature.


	"Apache logs (error_log) 'no acceptable variant' (406)) twice".

	Sorry the initial claim is wrong in general, not sure if this is
	a bug or a feature [robh] 

	DirectoryIndex index.html index.forward
	Multiviews switched on
	files: index.html.en index.forward
	browser set to accept say "fr".

	before index.forward (an .asis file) internally redirects the
	request to index.html a 'no acceptable variant' is logged,
	it is logged again when index.html fails to find an
	acceptable language.

	So what is apache looking for before the internal redirect from the
	.asis ?   If I remove index.forward from the equation, I get 1
	logged message plus a directory listing and a 200 response (yuck)

	SIGUSR1/graceful restarts 

	Problem: Under FreeBSD, this is lethal. USR1 followed by HUP takes down
	the server. Paul S (?) reported annoying processes after a USR1.

	Ben's offered a patch and it appears to cure the problem.
FIXED.	SSI causes bytes sent count to be waaay too low.

	For each SSI include, apache subtracts bytes sent so far from
	its counter (thinking bytes sent so far = headers only)

	patch accepted/committed


	From: Roy
	The problem is that a request on /dir/ is internally redirected to
	/dir/index.html, which properly results in a 304 Not Modified.  However,
	the r->status is not updated to reflect new->status (and anything else
	that might need to be promoted).

	we're still waiting for someone to speak up on whether it's a
	major problem or not.  I don't know that area of the code well
	enough, though I can't say that promotion of error code from
	sub-redirects has been a problem for me personally.  Again, flag
	it as a known bug is probably the best we can
	do here, unless someone wants to speak up.  [brian]


	1.2-dev still has this bug in it.  get_client_block needs to
	check for the -1 error case from the call to bread().

	patch offered. positive feedback so far.


	Roy's claim of CGI schitzo behavior and improper dealing with
	CONTENT_LENGTH.  This is a showstopper in my opinion until more
	voices are heard.

	Roy offers 2 options:
	  a) Assume it is only for old scripts.  This requires changing mod_cgi
	    so that it rejects anything without a Content-Length (411 Length
	    Required) even if it is chunked, or at least anything that cannot
            be read into a single buffer before execing the script.

          b) Assume it is for new scripts and that old scripts will just never
	    see the HTTP/1.1 input (after all, no sane client would chunk an
	    x-www-url-encoded form).  This requires changing get_client_block
	    so that it passes the chunk size and footer to the script. 

	Jim voted for (b)
	Ed  voted for (a) 

	Alexei says it's supposed to be this way.

	+ other comments.


done	SSI byte sent logging

	Roy's redirect bug, though he seems to be having 2nd thoughts now,
	probably not a showstopper.

	?  get_client_block needs to check for the -1  (sounds nasty)



done	ASAP	Find a patch for 'satisfy' that works well (tried and tested)
		 (before feature freeze)
		Prepare more documentation.
		Resolve all outstanding bugs (see above).
done	Nov 17	Feature freeze.
	Nov 24	Create last pre-beta tarball (Randy?)
		Test, test, test.
		Report back with positive as well as negative comments.
	Nov 30	Create 1.2b1 tarball.
	  "	Update CVS to reflect 1.2b1 release.
	Dec  1  Release 1.2b1.
	  "	Put documentation online if there is any.
	  "	Hand 1.2 management over.

View raw message