mod_rewrite/7369: mod_rewrite escapes QUERY_STRINGs where it shouldn't

[Greg Ubben <gs...@nsa.gov>]

>Number:         7369
>Category:       mod_rewrite
>Synopsis:       mod_rewrite escapes QUERY_STRINGs where it shouldn't
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    apache
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   apache
>Arrival-Date:   Tue Mar 06 18:40:00 PST 2001
>Closed-Date:
>Last-Modified:
>Originator:     gsu@nsa.gov
>Release:        1.3.19
>Organization:
apache
>Environment:
Solaris 2.7 on Sun Ultra 10, gcc -O2 (though any environment)
>Description:
Hello.  I believe fix PR#4734 in Apache 1.3.10 ("Make sure mod_rewrite escapes
QUERY_STRINGS on redirects") went too far, or was fixed in the wrong place.
It breaks a major web infrastructure service we use for directing "virtual"
URLs to actual URLs using mod_rewrite's DBM capabilities.  Here is our
basic mod_rewrite configuration, which runs out of a virtual host
(say, urn.bar.com):

RewriteEngine on
RewriteMap  urn     dbm:/urn/data/urn-map
RewriteMap  tolower int:tolower
# Map URN to URL if in the table and it's a type mod_rewrite can redirect.
# If extra path info given, replace the last component of the URL with it.
RewriteCond ${tolower:$1}	(.*)
RewriteCond ${urn:%1}		^((https?|ftp)://[^?#$|]+)$
RewriteCond %1|$2		^([^|?]+)(/.*\|)?(/[^|]*)
RewriteRule ^/([^/]+)(.*)	%1%3		[QSA,R,L]
# pass remaining special cases and unknowns up to a CGI to handle
RewriteRule ^/(.*)	/urn/cgi-bin/goto/$1  [NS,T=application/x-httpd-cgi]

For example, if the DBM table has in it:
	feedback	http://foo.bar.com/cgi-bin/feedback.pl

and we access this service via the virtual URL:
	http://urn.bar.com/feedback?to=gsu@bar.com;subject=test+this

it used to correctly translate this request and redirect it to:
	http://foo.bar.com/cgi-bin/feedback.pl?to=gsu@bar.com;subject=test+this

However as of Apache 1.3.10, it's turning the ; in the query string into
a %3b, causing the resulting access to fail because the final CGI can't
parse out the CGI parameters:
	http://foo.bar.com/cgi-bin/feedback.pl?to=gsu@bar.com%3bsubject=test+this

This problem still happens as of Apache 1.3.19.  I don't fully understand the
original problem report or fix, but I think mod_rewrite should be able to
pass an incoming query_string intact on thru to the rewritten URL that it
will be redirected to.  (Unless I'm missing something.)
>How-To-Repeat:
Can't give an actual URL as this is on a closed net.  See example in
description above.  Don't think this is related to the DBM stuff, so
the example could probably be simplified quite a bit.  Probably other
characters than semicolon (;) that are also a problem.
>Fix:
Sorry, mod_rewrite is too complex for me!  Maybe original fix was applied
in too general a case?  (This is a great module by the way -- exactly fit
the problem we were trying to solve!)
>Release-Note:
>Audit-Trail:
>Unformatted:
 [In order for any reply to be added to the PR database, you need]
 [to include <ap...@Apache.Org> in the Cc line and make sure the]
 [subject line starts with the report component and number, with ]
 [or without any 'Re:' prefixes (such as "general/1098:" or      ]
 ["Re: general/1098:").  If the subject doesn't match this       ]
 [pattern, your message will be misfiled and ignored.  The       ]
 ["apbugs" address is not added to the Cc line of messages from  ]
 [the database automatically because of the potential for mail   ]
 [loops.  If you do not include this Cc, your reply may be ig-   ]
 [nored unless you are responding to an explicit request from a  ]
 [developer.  Reply only with text; DO NOT SEND ATTACHMENTS!     ]