Re: [Fwd: HTTP 1.1 header in Apache 1.2] (fwd)

[Alexei Kosut <ak...@nueva.pvt.k12.ca.us>]

FYI.

-- 
________________________________________________________________________
Alexei Kosut <ak...@nueva.pvt.k12.ca.us>      The Apache HTTP Server
URL: http://www.nueva.pvt.k12.ca.us/~akosut/   http://www.apache.org/

---------- Forwarded message ----------
Date: Tue, 31 Dec 1996 15:48:50 -0800
From: Ryan Watkins <va...@dimensionx.com>
To: Alexei Kosut <ak...@ace.nueva.pvt.k12.ca.us>
Subject: Re: [Fwd: HTTP 1.1 header in Apache 1.2]


I'm not on the Apache Mailing list anymore, though I poke in and read
thru the archive occationally.


Alexei Kosut writes:

> On Tue, 24 Dec 1996, Chuck Murcko wrote:
> 
> > More trouble on the horizon, I'm afraid.
> 
> Blah. Can someone verify if this is true with Sun's Java tools as
> well, or if it's limited to Symantec's?

Its a problem with the 1.0.2 JDK from Sun.  I've seen it in the
Win95/NT version of the Sun JDK 1.0.2, and apparently its a problem in
the Solaris and Mac versions as well.  

It doesnt seem broken in any browsers I know of though.

I've tested it and does appear fixed in the latest 1.1 version of
Sun's JDK.  Most other JDK's (like the Linux JDK) are a port of 1.0.2,
and are probably broken as well, and will be for a while.

When I first ran into this problem with apache 1.2b1, I sent a note to
apache-bugs, and got a reply from Rob Hartill, and that there would be
no fix.  I've tweaked our server to just respond with HTTP/1.0 to
everything for now.  Our two main products at Dimension X are built in
Java and certainly not transitioned to the 1.1 (beta) JDK yet, so we
HAVE to work around this, and will probably have to keep some sort of
hack in for a good while.  I presume a number of other people are
going to run into the same problem.

Yea, the Sun JDK is screwed up, but ultimately our products have to
work for our clients, so its the server here we have to change.


> Although I'm defenitely a proponent of having HTTP/1.1 servers (like
> Apache) respond to HTTP/1.0 requests with HTTP/1.1, I'm beginning to
> wonder if it's really worth it. The only real *practical* reason to do
> it is so a client can determine whether a server supports HTTP/1.1 by
> sending a HTTP/1.0 request first to all servers. But it's very
> unlikely a client would do this - it'll most likely just send HTTP/1.1
> with all requests.
> 
> HTTP/1.0 clients that respond with an error to a HTTP/1.1 request
> aren't really broken (AOL is an exception, because they knew it would
> break things, btw), they're just pedantic. The HTTP/1.0 spec (as much
> as it is) doesn't really say anything about this, and nowhere does it
> explicitly state what should happen if it recieves a request of an
> HTTP version it doesn't understand. It's perfectly "correct" to
> declare it an error (as AOL did), or treat it as a HTTP/0.9 response
> (as apparently Cafe does), because it doesn't begin with "HTTP/1.0 ".

-- 
Ryan L. Watkins `silver moonbeams dance in fountains
                 below shining citadels
  vamp@vamp.org  surrounded by silver gates ascending silver stairs
   www.vamp.org  eureka on angelic prayer wafts in and scents the air' -satb