[NNTP] Future-proofing for including capabilities in responses

[Kai Henningsen]

clive at demon.net (Clive D.W. Feather)  wrote on 29.03.05 in <20050329080717.GA89808 at finch-staff-1.thus.net>:

> Ken Murchison said:
> > Then pick something that isn't likely to appear in the wild, e.g.:
> >
> > [:  :]  or [!  !] or <?  ?>
> >
> > Using non-ASCII seems overkill.
>
> Seeing as non-ASCII is very unlikely to appear in the wild, and
> illegal-UTF-8 even less likely, why isn't that better?

I consider specifying illegal UTF-8 a shooting offense. And I expect I'm  
not the only one. This is a MUST NOT, as far as I'm concerned - not  
negotiable.

I'm frankly appalled the notion has survived this long.

> We're talking about tokens that the client is going to examine; it's not
> that important that they be readable.

In any case, they shouldn't be invisible if all the rest of the protocol  
is plain visible text. That's just asking for trouble.

MfG Kai