Re: upstream and downstream

Doug Ewell wrote:
> If you have an application that lends itself to a limited repertoire,
> like automobile license plates or P.O. box numbers or house numbers, use
> that limited repertoire.  If you need a wider repertoire, you can use
> Unicode and you can *still* implement a subset.  The problem, as always,
> is in determining what the subset should be.

Well put, Doug. It is easy to see how stringprep and nameprep got to be 
the way they are. First, you want domain names in local languages. 
Second, you don't want to mess with the myriad charsets used around the 
world for something as central as DNS, so you choose the only real 
alternative, Unicode. Third, it is a lot of work to create a subset (and 
it's controversial), so you punt, and just publish the RFC anyway.

Didn't some CEO say, "Does it compile? OK, ship it!"

These comments are probably very upsetting to various people on this 
list. Sorry! All I'm pointing out is that it is easy to see how we ended 
up like this, human nature being the way it is.

Moreover, it is grossly unfair for someone like me to come in so late 
and make all these comments, when I could have had my eyes on the ball 
at the critical time, instead of going off and working on something 
totally different all these years. Then again, I might have missed it 
too, or just gone along with the flow, even if I *was* here. It's really 
hard to focus on the critical issues when you have an open mailing list 
with all sorts of people bringing up all sorts of little examples from 
around the world. I looked at the archives of this mailing list, and 
there is a lot of noise.

Erik