[hunchentoot-devel] Fwd: Intermittent problem with codequarterly website (chunked coding glitch?)
I have no particular reason to believe Hunchentoot is at fault here but I figured I'd ask in case anyone had seen anything like this. The codequarterly.com website is being served by Apache proxying to Hunchentoot (via ProxyPass and ProxyPassReverse directives in the Apache config) and this user reports occasionally getting back some hinky chunked-coded responses.
Any ideas of the top of anyone's head?
-Peter
---------- Forwarded message ---------- From: Thomas (AWS), Chris ccthomas@amazon.com Date: Thu, Jun 9, 2011 at 6:14 PM Subject: Intermittent problem with codequarterly website (chunked coding glitch?) To: "editor@codequarterly.com" editor@codequarterly.com
Greetings,
This may not be the appropriate place to report this issue, but I couldn't find any technical contact information on your website. Feel free to forward as necessary.
Intermittently (maybe 10-20% of the time), when trying to view pages on the codequarterly website, my browser is displaying what appears to be the text of the raw HTML of the page, instead of the rendered page. When I look closer, I see that the text begins with a hexadecimal number (e.g, '1FED'), followed by the usual HTML <DOCTYPE.... Header.
I reproduced the problem with curl & took a packet trace (output at end of email), to better see what is going on:
% curl -v http://www.codequarterly.com/2011/rich-hickey/
It shows that the response is coming back with chunked coding, but there are two hexadecimal numbers preceding the actual HTML content, instead of just the single number that one would expect from a chunked-coded response. The remainder of the body shows additional "extra" chunk length prefixes. It is almost as if the HTTP server was sending the response body through two passes of chunked encoding.
Anyway, you might want to look into how your website software is set up that might be causing this, or perhaps whether there's a proxy that is inappropriately adding its own chunked coding to the response as it passes through.
Best Regards, Chris Thomas
-- request -- GET /2011/rich-hickey/ HTTP/1.1 User-Agent: curl/7.18.1 (i686-pc-linux-gnu) libcurl/7.18.1 OpenSSL/0.9.8b zlib/1.2.3 Host: www.codequarterly.com Accept: */*
-- response -- HTTP/1.1 200 OK Date: Fri, 10 Jun 2011 00:51:12 GMT Transfer-Encoding: chunked Content-Type: text/plain
1f3e 1FE4<!DOCTYPE html> <HTML lang='en'> <HEAD> <META http-equiv='content-type' content='text/html; charset=UTF-8'> <TITLE>Code Quarterly</TITLE> ....
Any ideas of the top of anyone's head?
I think the problem is in any case would be solved if always specify the "Content-Length" and not use the "Transfer-Encoding: chunked"
Andrey
2011/6/10 Peter Seibel peter@gigamonkeys.com:
I have no particular reason to believe Hunchentoot is at fault here but I figured I'd ask in case anyone had seen anything like this. The codequarterly.com website is being served by Apache proxying to Hunchentoot (via ProxyPass and ProxyPassReverse directives in the Apache config) and this user reports occasionally getting back some hinky chunked-coded responses.
Any ideas of the top of anyone's head?
-Peter
---------- Forwarded message ---------- From: Thomas (AWS), Chris ccthomas@amazon.com Date: Thu, Jun 9, 2011 at 6:14 PM Subject: Intermittent problem with codequarterly website (chunked coding glitch?) To: "editor@codequarterly.com" editor@codequarterly.com
Greetings,
This may not be the appropriate place to report this issue, but I couldn't find any technical contact information on your website. Feel free to forward as necessary.
Intermittently (maybe 10-20% of the time), when trying to view pages on the codequarterly website, my browser is displaying what appears to be the text of the raw HTML of the page, instead of the rendered page. When I look closer, I see that the text begins with a hexadecimal number (e.g, '1FED'), followed by the usual HTML <DOCTYPE.... Header.
I reproduced the problem with curl & took a packet trace (output at end of email), to better see what is going on:
% curl -v http://www.codequarterly.com/2011/rich-hickey/
It shows that the response is coming back with chunked coding, but there are two hexadecimal numbers preceding the actual HTML content, instead of just the single number that one would expect from a chunked-coded response. The remainder of the body shows additional "extra" chunk length prefixes. It is almost as if the HTTP server was sending the response body through two passes of chunked encoding.
Anyway, you might want to look into how your website software is set up that might be causing this, or perhaps whether there's a proxy that is inappropriately adding its own chunked coding to the response as it passes through.
Best Regards, Chris Thomas
-- request -- GET /2011/rich-hickey/ HTTP/1.1 User-Agent: curl/7.18.1 (i686-pc-linux-gnu) libcurl/7.18.1 OpenSSL/0.9.8b zlib/1.2.3 Host: www.codequarterly.com Accept: */*
-- response -- HTTP/1.1 200 OK Date: Fri, 10 Jun 2011 00:51:12 GMT Transfer-Encoding: chunked Content-Type: text/plain
1f3e 1FE4<!DOCTYPE html>
<HTML lang='en'> <HEAD> <META http-equiv='content-type' content='text/html; charset=UTF-8'> <TITLE>Code Quarterly</TITLE> ....
-- Peter Seibel http://www.codequarterly.com/
tbnl-devel site list tbnl-devel@common-lisp.net http://common-lisp.net/mailman/listinfo/tbnl-devel
-
Andrey Moskvitin -
Peter Seibel