Hi,

Michael Shields reported in Debian bug http://bugs.debian.org/318686:
> When privoxy receives a page using Transfer-Encoding: chunked, and then
> de-chunks it for filtering, it changes the header to "identity".  This
> value of the header was included in RFC 2616 by mistake; the working group
> meant to pull it out.  See <http://purl.org/NET/http-errata#identity>
> and the http-wg mailing list archives for discussion of this.
>.
> Specifically, this breaks the libneon that is bundled with vfs2,
> and therefore gweather-applet.  I am reporting a bug (#318683) against
> Privoxy also, with a patch to cause it not to send this header.  However,
> I believe that libneon is also at fault for assuming that any T-E header
> implies the chunked encoding it asked for.  The attached patch makes
> vfs2 tolerant of code that implemented the RFC as published.

He proposed the following patch:
--- gnome-vfs2-2.10.1/imported/neon.old/ne_request.c    2005-01-30
11:59:55.000000000 +0000
+++ gnome-vfs2-2.10.1/imported/neon/ne_request.c        2005-07-17
01:43:26.000000000 +0000
@@ -480,14 +480,18 @@
     return (st->klass == 2);
 }
 
-/* Handler for the "Transfer-Encoding" response header: treat *any*
- * such header as implying a chunked response, per the "Protocol
- * Compliance" statement in the manual. */
+/* Handler for the "Transfer-Encoding" response header.  Previous
+ * versions treated *any* such header as implying a chunked response.
+ * However, we need to handle at least "Transfer-Encoding: identity"
+ * also (by ignoring it).  This was included in RFC 2616 by mistake
+ * (see http://purl.org/NET/http-errata#identity), but it is there,
+ * and Privoxy uses it at least through version 3.0.3. */
 static void te_hdr_handler(void *userdata, const char *value) 
 {
     struct ne_response *resp = userdata;
-
-    resp->mode = R_CHUNKED;
+    if (strcasecmp(value, "chunked") == 0) {
+       resp->mode = R_CHUNKED;
+    }
 }

 /* Handler for the "Connection" response header */


Bye,