Re: [remote protocol] support for disabling packet acknowledgement

[Paul Koning <Paul_Koning@dell.com>]

>>>>> "Pedro" == Pedro Alves <pedro@codesourcery.com> writes:

 Pedro> A Friday 11 July 2008 16:53:46, Paul Koning wrote:
 >> >>>>> "Daniel" == Daniel Jacobowitz <drow@false.org> writes:
 Daniel> On Fri, Jul 11, 2008 at 04:10:47PM +0100, Pedro Alves wrote:
 >> Agreed.  Telnet to a terminal server that feeds a UART based
 >> target stub is common practice.

 Pedro> Ack, we're all in sync.

 >> TCP at both ends with datagrams in between is too weird to
 >> consider.

 Pedro> Not weird at all, and it is safe.  It doesn't matter what you
 Pedro> have in the middle as long as both ends have tcp.

You're probably thinking about end to end TCP over a datagram cloud.
That works, of course, that's the Internet.  I was talking about TCP
from A to B, raw UART B to C, TCP from C to D.  TCP wouldn't be
helping you detect or correct data loss on the B to C path, and that
means you'd need application layer acks (as in the current remote GDB
protocol) for that case.  But that's a topology that makes no sense to
me and I wouldn't expect ever to see.

   paul