ATTN: Developers of NCSA & CERN HTTP servers

"William M. Perry" <wmperry@cs.indiana.edu>
To: www-talk@nxoc01.cern.ch
Subject: ATTN: Developers of NCSA & CERN HTTP servers
Date: Tue, 27 Jul 1993 14:32:56 -0500
Message-id: <13917.743801576@spinner.cs.indiana.edu>
From: "William M. Perry" <wmperry@cs.indiana.edu>
Status: RO

Just wanted to take a minute to point out a few bugs in the servers
put out by NCSA & CERN.  It doesn't really affect most browsers (yet),
but it might bite people who start using HTTP/1.0 aware browsers
(lynx, and my Emacs browser are the only two I am aware of at the
moment)

Problem is - when sending multiple lines of an HTTP/1.0 request to a
HTTP/.9 server, the output gets truncated if the lines are sent too
fast.

For example, type this in an emacs buffer:

(let ((proc (open-network-stream "WWW" "WWW" "www.ncsa.uiuc.edu" 80)))
  (process-send-string proc "GET /EVL/docs/Welcome.html HTRQ")
  (process-send-string proc "From: someuser@somesite"))

Then evaluate it - you will get a buffer called WWW that has only
about 1/2 of the text of the EVL Welcome page.

You can get the same effects outside of emacs by cutting and pasting
in an Xterm.  

> telnet info.cern.ch 80
GET / HTRQ  	    	    	    \_ Cut & paste in xterm
From: someuser@somesite	    	    /  typing it by hand is too slow

and you will get a truncated version of the CERN home page.

I was able to get this to appear in lynx (which uses the new library
from cern for HTTP/1.0) to do this - it took a _long_ time though.  I
guess the cern library spaces out its request lines when sending to
the server.

Does anyone know what could be causing this?  I've gotten around it
for now in my emacs browser, but it was a pain tracking it down - I
thought it was a bug in emacs, and I asked rms for help and then he
discovered it wasn't emacs fault.  (Hallelujah, praise the one true
editor :)

As I said, I've gotten around it for now, and hopefully HTTP/0.9 only
servers won't be around much longer, but it would be nice if this was
fixed, or at least documented so you could call it a feature. :)

Thanks,
    Bill Perry