Re: cvs commit: apache/src buff.h buff.c http_main.c http_protocol.c

[Marc Slemko <ma...@znep.com>]

On Sat, 25 Jan 1997, Roy Fielding wrote:

> fielding    97/01/25 14:37:17
> 
>   Modified:    src       buff.h buff.c http_main.c http_protocol.c
>   Log:
>   Fixed bug in bcwrite regarding failure to account for partial writes.
>   Avoids calling bflush() when the client is pipelining requests.
>   Remove unnecessary flushes from http_protocol.
>   
>   Submitted by: Dean Gaudet
>   Reviewed by: Roy Fielding, Alexei Kosut
>   
>   --- 1675,1684 ----
>             if (r) increment_counts(child_num,r,1);
>     #endif
>     	while (r && current_conn->keepalive) {
>   ! 	    /* If there's no request waiting for us to read then flush the
>   ! 	     * socket.  This is to avoid tickling a bug in older Netscape
>   ! 	     * clients. */
>   ! 	    if( btestread(conn_io) == -1 ) bflush(conn_io);
>     	    destroy_pool(r->pool);
>     	    (void)update_child_status (child_num, SERVER_BUSY_KEEPALIVE,
>     	     (request_rec*)NULL);

Is this one really only for broken Netscape?

Isn't it just a fact of life that if there is no request coming in, stuff
could wait around in our buffer until it is closed if it isn't flushed?