The only way to guarantee QOS is total control over all elements in the connection. Other schemes are fabrication/dreaming. I could recount some *clever* attempts but they are too humourous for credibility - though they were impressive to some (management) and kept people in jobs.

Cheshire's laws of Network Dynamics:
http://www.stuartcheshire.org/rants/Networkdynamics.html

1. A guaranteed network service guarantees to be a low quality overpriced service

2. For every guarantee there's a corresponding refusal

3. For every Network Connection there's a corresponding Network Disconnection

Last I heard, (a year or two ago when I did this stuff,) what the big vendors mean by QoS is that you configure a router to prioritize traffic by IP or by protocol. No cooperation from the end-points is required.

Of course it does not completely guarrantee actual quality of real service, but it's still useful. Not antithetical to open standards, but potentially dangerous for the end-to-end principle...?

This speaks to 'what is QOS?' It is not the same as 'complete control over the experience.' Or maybe it is :-) Maybe it is like Perl Taint checking...you are tainted if you touch a 'non qos certified' bit of kit.

But this is irrelevant to my concerns, because except in very specific cases which to date I find uniteresting, I dislike the whole notion of QOS when applied to general purpose networks (ie. If you run fibre between all of your buildings world wide, then you can worry about QOS, and it is possibly vital for you operations, but in the Big I Internet...no way).

QOS stands in oppostion to the 'worse is better' 'greasy kids' non-optimized network that we know and love.

Seems that if you really want QoS you need to include in your standard some way to query the other end and the stations in between as to what they will do for you. Then, to really think outside the box here, you could pay them for giving you a certain guaranteed level of service, and sue them if they didn't provide what they promised. This would allow both completely open implmentation of standards, and a mechanism for people to be given a binding promise of a certain quality of service.

Open standards actually solve these sorts of problems. If you've got a fixed protocol and those authorised to exchange data with you in this protocol start to suck or disappear, you're screwed.
On the other hand, if it's an open protocol, then if a remote provider starts to suck, I can decide not to do business with them anymore. I'll switch to another provider. Since anyone can use the protocol without paying huge licensing fees, then there is more likely to be more providers.
If all the providers for a certain data type suck, then it doesn't really matter whether the protocol is open or closed.