From: Russ Allbery (rra@stanford.edu)
Date: Tue Oct 22 2002 - 11:49:28 CDT
Jean-Marc Desperrier <jean-marc.desperrier@certplus.com> writes:
> Russ Allbery a dit :
>> Those of you who participate in national hierarchies where this is
>> really important have an excellent testing bed to start working on a
>> solution, much better than the rest of us do. You have a limited
>> audience, often on a reasonably defined set of servers, you can
>> determine the major clients the local users are using to read news, and
>> you can try things locally without requiring anyone else's approval.
> The trouble is that this what is happening now, and what has been
> implemented since _years_, and that the result is a solution that works
> only locally, on a limited audience, on local user.
You have non-ASCII newsgroup names right now in national hierarchies?
I don't think that the problem has actually been completely solved even
locally. I also don't think that having a multitude of character sets is
as big of a problem as people seem to think, but that's mostly based on my
judgement of the behavior of the most commonly used clients and may well
be wrong.
> The problem is that if you want to normalize and universalize it, it
> _does not work_, because the local encoding is never the same
> everywhere.
If this doesn't work, then it causes some sort of drawback, some sort of
problem that people want solved. If it causes a problem that people want
solved, there's an opening to come up with a better solution. If there
isn't any problem that people want solved enough to adopt a different
piece of software or a different implementation to solve it, the problem
isn't worth solving.
Again, this is an argument from pure pragmatism. At some point, someone
has to either do the work or pay for the work to be done. Unless you can
find a problem annoying enough and a solution compelling enough to get
someone to do that, the problem will remain unsolved, no matter how much
we talk about it.
> What USEFOR must do is to set something that intead of working only for
> a limited subset really works universally.
> And we had almost reached that two months ago, before a bunch of new
> people arrived, went into endless arguing,
No, USEFOR never had a consensus. The voluable people preferring a
particular solution just kept talking to the point that many of us gave up
debating it endlessly.
> No the issue is not fundamentally hard.
I really can't agree with that.
> And the difficulty is not to write new code that support the new
> protocol, but to get the owners of popular programms to update their
> code to support the new format.
Yes, we've heard that before with other things, but the easy-to-write code
never materialized.
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>