[mailto:owner-ietf-mxcomp(_at_)mail(_dot_)imc(_dot_)org] On Behalf Of Dean
Anderson
Sure, the consensus among network engineers and scientists.
Sure they aren't typical of internet users. Phone engineers
are atypical of phone users. Auto engineers are atypical of
auto drivers, too. But who do you want making decisions at
the Department of Transportation or running the engineering
department at a car company?
I think it is ok for us to take decisions for others, but we have to be
willing to understand that their needs and concerns are not the same as
ours.
There is a difference between a prototype and a product. Unfortunately we
have a habbit of inflicting PERL lash up style schemes that are really no
more than prototypes onto end users.
I spend too much of my time dealling with the technology problems of friends
and familly. RTFM is not an acceptable answer here, there should never have
been the need for a manual. The only manuals I ever read are the ones that
come with safety warnings, things like tables saws, joiners, planers etc.
I'm not so concerned about the selection of decision makers
and leaders from the technical community, so much as I'm
concerned about whether the decision makers selected actually
represent the technical community, rather than a certain
small group in the community.
I am starting to get to a clearer explanation of the perception gap I see in
standards making.
When writing a standard people imagine that they are writing the rules. The
problem is that the people who read the standards are looking to find WHAT
WORKS, not what is right.
I do not do normative ethics, you cannot arrive at an ought from an is.
Regardless of what people imagine SHOULD be the situation the fact is that
the way standards documents are used is very different from the way people
writing them imagine.
If we want to get the implementers to produce stuff that complies with the
specs we have to make it so that what works and what is right are the same
thing. Test suites help close that gap. Designing specs for compatability
with actual legacy deployments is another part.
The process is somewhat opaque alright. And it resembles
more of a private club than an public service organ or NGO or
standards body. But nominations imply elections however
informal or exclusive the consensus is.
Open and inclusive implies open and inclusive elections.
There is a fundamental concept of democracy and consensus to
the IETF. Its just insufficient and flawed. Most standards
organizations have a membership concept, and /all/ members
get to vote on certain things.
If someone wants to trump my opinion in my field of expertise on an issue I
believe is critical to the security of the Internet on the basis of an
unelected appointment through the old boy club then they are going to have a
big fight on their hands.
I find it utterly ludicrous that we are on the brink of taking a spec to an
alternative working group due to a technical dispute where the pragmatic
approach is so clear.
If people don't want a fight then the answer is clear, give me an
alternative way to deploy MASS that does not create a dependency on the
deployment of new DNS infrastructure. The dependency is real and it is
neither necessary nor even advantageous.
Right. "Combining relative weights" sounds very much like Bayes rule.
But Bayesian inference makes some assumptions that are not valid for a real
corpus, the probabilities are not disjoint and Bayes developed his theory of
probability for that area. It's a bit like saying Einstein's cosmology is
Euclidean because on almost any terrestrial measurable scale it approximates
to the same thing, but the whole point is that Einsten's cosmology is
non-Euclidean.