Index: server/core.c
===================================================================
--- server/core.c	(revision 1042903)
+++ server/core.c	(working copy)
@@ -73,6 +73,9 @@
 #define AP_MAX_INCLUDE_DEPTH            (128)
 #endif
 
+/* valid in core-conf, but not in runtime r->used_path_info */
+#define AP_ACCEPT_PATHINFO_UNSET 3 
+
 APR_HOOK_STRUCT(
     APR_HOOK_LINK(get_mgmt_items)
 )
@@ -120,7 +123,7 @@
     conf->override_opts = OPT_UNSET | OPT_ALL | OPT_SYM_OWNER | OPT_MULTI;
 
     conf->content_md5 = 2;
-    conf->accept_path_info = 3;
+    conf->accept_path_info = AP_ACCEPT_PATHINFO_UNSET;
 
     conf->use_canonical_name = USE_CANONICAL_NAME_UNSET;
     conf->use_canonical_phys_port = USE_CANONICAL_PHYS_PORT_UNSET;
@@ -3759,15 +3762,17 @@
     /* Deal with the poor soul who is trying to force path_info to be
      * accepted within the core_handler, where they will let the subreq
      * address its contents.  This is toggled by the user in the very
-     * beginning of the fixup phase, so modules should override the user's
+     * beginning of the fixup phase (here!), so modules should override the user's
      * discretion in their own module fixup phase.  It is tristate, if
-     * the user doesn't specify, the result is 2 (which the module may
-     * interpret to its own customary behavior.)  It won't be touched
-     * if the value is no longer undefined (2), so any module changing
-     * the value prior to the fixup phase OVERRIDES the user's choice.
+     * the user doesn't specify, the result is AP_REQ_DEFAULT_PATH_INFO.
+     * (which the module may interpret to its own customary behavior.)  
+     * It won't be touched if the value is no longer AP_ACCEPT_PATHINFO_UNSET,
+     * so any module changing the value prior to the fixup phase 
+     * OVERRIDES the user's choice.
      */
     if ((r->used_path_info == AP_REQ_DEFAULT_PATH_INFO)
-        && (conf->accept_path_info != 3)) {
+        && (conf->accept_path_info != AP_ACCEPT_PATHINFO_UNSET)) {
+        /* No module knew better, and the user coded AcceptPathInfo */
         r->used_path_info = conf->accept_path_info;
     }
 
Index: modules/http/http_request.c
===================================================================
--- modules/http/http_request.c	(revision 1042903)
+++ modules/http/http_request.c	(working copy)
@@ -521,6 +521,11 @@
     apr_table_setn(new->subprocess_env, "REDIRECT_STATUS",
                    apr_itoa(r->pool, r->status));
 
+    /* Begin by presuming any module can make its own path_info assumptions,
+     * until some module interjects and changes the value.
+     */
+    new->used_path_info = AP_REQ_DEFAULT_PATH_INFO;
+
     /*
      * XXX: hmm.  This is because mod_setenvif and mod_unique_id really need
      * to do their thing on internal redirects as well.  Perhaps this is a