> --On Tuesday, December 3, 2002 8:10 AM -0800 Brian Pane > <[EMAIL PROTECTED]> wrote: > > >> This sounds a bit more reasonable to me. That is, send chunked if > >> the client will accept chunked, else send a connection: close > >> header (which will tell the client we are done sending). As a > >> compromise, we could start off buffering and if we hit some magic > >> threshold and we still do not know the c-l, add a connection: > >> close header and start sending. > > > > +1. Buffering up to, say, 8KB sounds good to me. For anything > > larger than that, there's little harm in sending Connection: close. > > Actually, I just remembered that's invalid. > A request can't indicate > its end of the request body with a Connection: Close. A server can > do that for a response, but not a client with a request. I remember > encountering this situation before (with a really poorly implemented > custom proxy), and Roy pointed out that this is a no-no in the RFC. > (You can't get away with a 'half-close.') > > So, it's either T-E or C-L. If it is C-L, we must buffer everything. > No way around this. -- justin >
Oops, right you are... I just re-read your original message. I missed the fact that you were talking about a -request- not a response. Bleh. Bill
