ietf-mxcomp
[Top] [All Lists]

RE: Status of MARID WG?

2004-11-17 12:29:13

All you need is two interoperable implementations and a userbase 
noticable to the IETF. The fact that 98% of the users of 
the internet 
will never use it directly or indirectly does not matter.

The "userbase noticable" isn't requirement of the RFC 
process. The IETF 
doesn't standardize based on popularity, but on consensus.

Consensus amongst a community that is 100% atypical of the Internet users.

Each time I vist the IETF the proportion of members who are unashamed
technologist supremacists rises. There is no embarassment about developing
systems that ordinary users cannot make use of.



Nobody in the IETF is elected, nobody is accountable. The 
inevitable 
consequence of that situation is that nothing that the IETF 
does can 
ever rise above the level of a personal opinion.

This is also not true.  It may appear this way from time to 
time, but it isn't literally true.

It is literaly true that nobody is elected. All appointments are through a
selection committee that is explicitly non-representative and
non-accountable.


Indeed, the Microsoft presentation given at the Anti-spam 
conference at MIT demonstrated that.  The MSN people noted 
that their biggest complaints came from other divisions of 
Microsoft whose spam they blocked (and that was unblocked). 
And while they weren't sharing their techniques (better to 
exploit them that way), what little they did reveal made me 
think they were using bayesian technique, though I think they 
specifically denied using spam-bayes. I thought it odd for 
them to come to a technical conference, give a presentation, 
and not actually share any technical information.

The best presentation at the MIT conference was by the MIT undergrad who was
the only person there who gave a rundown of the comparative effectiveness of
bayesian inference vs other techniques and found the other techniques worked
better.

I would be surprised if Microsoft did use Bayesian inference, they are not
the best tool for their corpus by a very long way. In the presentation I
heard the presenter said that they used a huge number of rules and
constantly re-evaluated both the ones that were most effective and the ways
in which relative weights were combined.


These same people saying spam-bayes is pure evil (eg 
Vixie) have made statements to the effect that anything that 
helps reduce spam is good.  Except spam-bayes. They say 
that's bad. Not just a little bad. But terribly, horribly, 
drastically bad.

Oh I can think of ways in which Spam-bayes might be bad for Paul Vixie...


I've done some work applying information theory to spam, and 
have discovered that there is no ultimate solution to spam. 

Now there is an interesting statement. Whayt if the solution does not exist
in information theory?

But 
there is a nomination process for the IETF chairman, and 
members of the IAB.  The IETF is also an activity of ICANN, 

No it isn't.

And no, a nomination process does not an accountability mechanism make.


<Prev in Thread] Current Thread [Next in Thread>