Lists Home |
Date Index |
On Wed, 23 Jan 2002, David Starr wrote:
> In the abstract, I like the notion that a better real-time response allows
> finer grained transactions. I can see how moving to UDP helps in the
> expected case, and maybe optimizing the expected case is sufficient for
> most. However if true real-time behavior is required, don't you need a
> connection oriented protocol to ensure QoS?
You don't need a connection oriented protocol per se, but you do need
*some* notion of connection *somewhere*. In some such protocols you merely
agree with the network that all packets marked with a certain flag between
you and a specified remote host will have certain guarantees of delivery
within a certain timeframe as long as the bandwidth they attempt to use
doesn't go above the agreed limit.
This 'flow', as it is called, can then be used for any packets you want -
TCP, UDP, etc...
Alaric B. Snell
http://www.alaric-snell.com/ http://RFC.net/ http://www.warhead.org.uk/
Any sufficiently advanced technology can be emulated in software