On Mon, May 31, 2010 at 09:16, Qoo Goo <qoo...@gmail.com> wrote: > Hi Derrell and everyone, > > We've been able to solve the problem. It was one of those kind of > extremely silly problems that sometimes are hard to find. I post the > explanation here just to prevent others (we've lost almost two days of > work!). > > I was just that we were closing the socket in server side. I don't now > why, but in one of our server's scripts we were erroneously calling > (asp.net) Response.Close() instead of Response.End() after flushing > contents to the client. > > The first method closes the socket and it seems that Google Chrome 5 > (only Chrome 5) doesn't like it.
It's not really a "doesn't like it" scenario, it's just a difference of implementation. Both are correct. There is no guarantee that when the server closes the socket, that all data that the server has sent has yet been received by the peer (client), let alone processed by the client. Most likely (but not necessarily), if data is sent before a connection is closed, that data arrives at the client and is queued for processing at the client. If it has not yet been processed at the time that the client receives the connection reset (closing), the client has the option of immediately flushing the data queue, or of processing what's left in that queue and then processing the close operation. Apparently, Chrome is choosing to flush the queue immediately upon receipt of the command to close the stream, whereas the other browsers are queuing the close command behind the pending data commands, so they are processed in the order received. TCP supports a "negotiated close" operation, which I'm guessing is what Response.End() initiates. That would ask the peer to agree to the close, which it wouldn't do until it had processed the pending messages. I'm guessing that Response.Close() does an immediate reset of the connection, and since Chrome is not queuing that request you lose your connection immediately. So... do the right thing. :-) Call Response.End() and you won't have the problem. Derrell
------------------------------------------------------------------------------
_______________________________________________ qooxdoo-devel mailing list qooxdoo-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/qooxdoo-devel
