Re: [precis] LetterDigits

[Peter Saint-Andre <[e-mail]>]

On 4/10/12 1:30 PM, Andrew Sullivan wrote:

> Ok, surely we need a class between NameClass (which I grant is too
> restrictive for a lot of things) and FreeClass 

It all depends on what our customers want. To date, other than our
customers for SASL user names (discussed in Paris) and possibly (*) LDAP
distinguished names, we haven't seen demand for a string class between
NameClass and FreeClass.

* I say possibly because we really haven't received any feedback at all
from our LDAP customers, which worries me.

> (which I guess ought to
> be as open as possible, to account for the possibility of
> subclassing). 

Right. If you want something less free than the FreeClass, subclass it.

> I know we had many more of these in the past, 

We had one more: the SecretClass for passwords. However, we decided to
merge that into the FreeClass because it seemed to be a distinction
without a difference.

> and
> ditched a lot of the distinctions, but can't we come up with
> NonBoneheadClass that restricts us to the set of things that can be
> predictably normalised in some sane way?  We can't solve the O -> o
> and ò -> Ò problem, but surely we could set a class that doesn't
> include those cases?  Yes, this might entail nasty mappings and
> information loss; that would be the point.

Do our customers want or need that string class?

Peter

-- 
Peter Saint-Andre
https://stpeter.im/