--- bill lam <[EMAIL PROTECTED]> wrote:

> Oleg Kobchenko wrote:
> > Yes, but HTTP/1.1 uses chunks, which instead of Content-Length
> > prepends each chunk with count. So the byte counting is
> > left for the next iteration. It should have separate
> > layers for reading sockets and processing 
> > content: headers and chunks. It has to be buffered, as chunks
> > may not coinside with boundaries of blocks.
> 
> chunk in HTTP/1.1 is really troublesome, thus is there any good reason to 
> send a
> HTTP/1.1 request?
> 
> Most servers will accept HTTP/1.0 request, did you encounter any server that 
> do
> not honour HTTP/1.0 request?
> 
> For the server side, it is legal to fallback to HTTP/1.0 when receiving 
> HTTP/1.1
> request.

Yes, but HTTP/1.1 is there for completeness. Chunk processing is done anyway.
And the fact that HTTP/1.1 may not have Content-Length can have the same effect
as HTTP/1.0 without thereof.

> Implementing a reliable client need a state machine because packets can arrive
> in any manner, it has to prepare for the case where data arrive one byte each 
> time.

But for the purpose of getting Content-Length and identifying the beginning and
end of body as simple two level solution can do.



__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to