From: Bruce Lilly (blilly@erols.com)
Date: Thu Sep 26 2002 - 23:02:33 CDT
Charles Lindsey wrote:
> In the cases where there is the possibility of using RFC 2047/2231, it is
> always safe because the implementor of the posting agent can make the
> necessary arrangements. If he chooses to encode every time, then that is
> fine. If he chooses to stick with UTF-8 wherever possible, that is fine
> too.
Precisely how does the implementor of a posting agent control
the implementation of injection agents and gateways? Or, for
that matter, user agents which are not themselves posting agents?
> And the point with Newsgroups is that
> whatever we decide now will be fixed FOR EVER (since it has to be a
> canonical form on the wire, and it will be impossible to change to a
> different canonical form later on).
That argument doesn't ring true -- if it will be impossible to
change later, surely by the same rationale (whatever that might be)
we cannot change now; or conversely, if we can change representation
now, we could as easily do so at a later date. For example, we could
now extend newsgroup name components to include =XX, which is
compatible with what is likely to remain in place for another decade
at least, and possibly further extend it at some time in the future
when conditions are appropriate.
> In 20 years time (and possibly in 2 years time), any canonical form that
> still tries to cram 8bit codes into 7bits is going to look totally and
> utterly stupid.
To paraphrase Keith Moore, please don't quote an extrapolation of
20 years of current practice as fact.