On 9/22/05 1:14 AM, "Dave Crocker" <dhc2 at dcrocker.net> wrote:
> The term
> "real-time" tends to mean sub-second, and often much faster than that.
That seems to be the vernacular use, but strictly speaking "real-time"
is about robust assurances of delivery within a constrained time period,
whether that time period is millisecond or multi-minute. In other
words, the focus is on the guarantees, not on really-really-fast.
Let me play Devil's Advocate for a second: What wasn't addressed by the
previous work done in RSVP and Diffserv that this new area has to be
urgently addressed? Are not packet schedulers enough?
Or are we going to watch a whole new infrastructure get developed that
will be now better than the previous work, with less complexity, and
more hope of being deployed outside of DoD programs? Because, frankly,
that's the only place I've seen an explicit need for advanced QoS
techniques (besides VoIP, and that gets to be really debatable too.)
-scooter
