How the heck are people able to deploy native-IP networks
with these kinds of limitations/problems with QoS? Or did I
miss something about QoS recently?
Yep. The biggest QoS secret is that nobody actually needs
it. Bandwidth is cheap and is growing cheaper. The
manpower needed to deploy and maintain QoS is getting
more and more expensive.
Overbuild strategies are the most common from what I've seen, but
are also fragile. Technological history is replete with "overbuilds"
that either ran into their end of life too soon, or were pressed
into service far longer than the designers intended. [insert
obligatory mickeysoft memory management flame-by-example]
I'll agree that rather than concentrating on the latest and greatest
QoS "if everyone adopted <foo> we'd be fine" strategies (that sound
identical to the Promise of an ATM World), we should be concentrating
on things that have been demonstrated to improve quality (CoS-based
queueing, modern queue management, intra-domain CoS-tags, replacement
of UDP by TCP wherever possible, content replication/caching, etc).
The issue of adapting your CoS-queues -across the domain- to demand
seems to me the sharp edge of real VoIP deployment.