Re: POST forms & Netscape Navigator

Addressed to:

** Reply to note from Matthew Denner <> Thu, 25 Jul 1996 12:10:01 +0100 
> > > Instead of reading until the end of the connection you should find the  
> > > "Content-Length" header which tells you the size of the POST'ed data.  
> > > Read this number of bytes after the 2 CRLF's.  
> I do read the Content-Length header and I do wait for the correct number  
> of  
> bytes to come through.  However, the connection has no data present for  
> collection and the machine simply sits there waiting.  
Have you tried using select on the socket to implement a timeout 
at least? 
> > to read. Doing the keep-alive is optional if you're doing a HTTP/1.0  
> > server. It's not part of the "standard" part of RFC 1945. It is the default  
> > for HTTP/1.1 which means as a server you must always send a valid Content-Length  
> > when sending an entity body to the client.  
> >   
> The keep-alive header isn't implemented properly in Netscape as far as I  
> can  
> tell.  The connection is continually closed by Navigator after just one  
> request.  
> But I'm not sure on this as it could be my programming.  
Don't know about this one. My HTTP/1.0 server doesn't do persistent at all. 
> > A technical point here also (probably minor and fussy), but the way you  
> > view the separation between headers and entity body is there is only  
> > one empty line (CRLF). The first CRLF is attached as the end marker for the  
> > last header. The separation is then marked by an emtpy line. Whatever comes  
> > after that is the entity body.  
> >   
> There are two CRLF's: the last line of the header, and the empty line.   
> Besides  
> which I have altered the code to check for an empty line. 
What I'm saying is the CRLF after a header is considered part of the header, 
not part of the next line. Something like: 
  POST /cgi-bin/test HTTP/1.0CRLF 
  Content-Type: text/plainCRLF 
  Content-Length: 50CRLF 
  post data starts ... 
At least from my perspective, the CRLF is logically part of the header 
line it is on. What ends the headers is a line without data, just CRLF. 
I guess later today, I'll have to do some testing with Netscape. I've 
just used my own testing program to date since I don't think Netscape 
implements any 1.1 stuff yet. It's also a heck of a lot easier to simulate 
when I can control the input exactly. 

Paul Hethmon 
Computerman -- Agricultural Policy Analysis Center 
NeoLogic Ftp & Mail Servers 
Knoxville Warp User's: 

Received on Thursday, 25 July 1996 08:33:15 UTC