BaseHTTPServer & TCP Reset

[Scherer, Bill]

On Thu, 24 Jan 2002, Steve Holden wrote:

> "Scherer, Bill" <Bill.Scherer at VerizonWireless.com> wrote in message
> news:mailman.1011884317.31625.python-list at python.org...
> >
> > I'm having a problem with an application using BaseHTTPServer.
> >
> > Certain results coming from one method in the application will
> > cause a reset at the tcp level.  This is causing great headaches,
> > for in IE 5.0, the user is simply warned of the connection reset
> > and left at that.
> >
> > IE 5.5 however, upon receiving the reset, will rePOST the
> > request.  Soetimes it even does a third resubmit, only as a GET
> > request.  !?!?!?!?!?
> >
> > The behavior is repeatable.  The response data is html, with
> > proper headers, content-length, etc.  Python is v2.1.1, and OS is
> > RH Linux 6.2.
> >
> > Any help is greatly appreciated.
> >
> > Thanks!
> >
> It's probably line 3 of your method terminating the output early :-)
> 
> A little more context might be helpful in debugging this problem. Are you
> absolutely POSITIVE that the Content-Length: header is correct? I've seen
> this odd behavior from IE in the past, and ISTR it was associated with
> output length.
> 
> Anyway, perhaps if we needed our psychic powers less, we could be more
> helpful.

Yeah, I expected that.  But at least I've got the conversation 
started...

I am fairly certain the content length is correct; all app 
methods respond via the same code.  Only the html from this one 
particular method has problems, and only for certain inputs.  
When it does happen, a view source on the client (before the auto 
re-submit) shows the complete source.  Sometimes the client sits 
there for a minute or two before it resubmits ( a wait cursor 
whilst hovering over the titlebar is an indicator that it will 
re-submit in the near future). 

Also, I've tried it without setting content length at all (if i 
read the spec correctly, that's allowed; the client is supposed 
to know that it's done when the server closes the connection), 
and that didn't change the behavior.

I've tried alter my respond code in many ways, with waits and 
flushes, but nothing seems to help.  Here's what it looks like:

def respond(self, html, ctype='text/html', fname=None):

        self.length = len(html)

        baseType = ctype.split('/')[0]
        
        if (baseType == 'text') and (type(html) != types.StringType):
            html = repr(html)

        expiration = time.strftime("%a, %d %b %Y %H:%M:%S GMT", \
                     time.gmtime(time.time()))

        try:

            self.send_response(200)
            self.send_header("Content-Type", ctype)
            self.send_header("Content-Length", self.length)
	    self.send_header("Expires", expiration)
	    self.send_header("Connection", "Close")

            if fname:
                self.send_header("Content-Disposition", \
                                  'inline; filename=%s' % fname)

            self.end_headers()

            self.wfile.write(html)
	     
            time.sleep(.1)
            self.wfile.flush()

	    
        except IOError:
            #client connection has gone away... 8-(
	    pass




Again, I'm reponding with this code from all the other methods 
without issue, send csv's, pdf's, and html.  Given that I'd 
suspect the actual html from the instances where it goes wrong, 
but I've looked at that, and I can't see anything that would 
cause a server reset...




> 
> regards
>  Steve
> --
> Consulting, training, speaking: http://www.holdenweb.com/
> Python Web Programming: http://pydish.holdenweb.com/pwp/
> 
> 
> 
> 
> 
> 

-- 
Bill.Scherer at Verizon Wireless
RHCE 807101044903581