On 2/3/15 10:18 PM, Bjoern Hoehrmann wrote:
* Pete Resnick wrote:
Your suggestion I think is making too strong a claim, but I see where
you're going. You needn't limit to a single script or otherwise heavily
restrict to stay clear of potential problems; you simply have to stick
to the more restrictive classes provided, or if you need to use
free-form, then restrict to something more limited. So perhaps this
would be clearer, and capture your concern:
Even so, implementations that are sensitive to the advice given in
this specification (to use the more restrictive String Classes, or
otherwise to only allow a restricted set of characters, particularly
ones whose implications they actually understand) are unlikely to
run into significant problems as a consequence of these issues or
potential changes.
This is better, but if I and my implementation do run into significant
problems because I did not disallow some characters that I could have
disallowed, do you really think that I am to blame? I18N folks tell me
to have as few restrictions on characters as possible, while the text
above suggests to have as many restrictions as possible. If I fail to
find the proper balance myself, in the void that is left by the draft,
I think the specification is also responsible.
I'm not sure who these i18n folks are who are telling you to have as few
restrictions as possible. The very point of the PRECIS work, in part,
has been to define a restricted, safe set of characters that is still
expressive enough for most humans - i.e., the IdentifierClass. (The
FreeformClass goes in the opposite direction, but then we provide some
strong warnings about using that.)
Yes, the balance between safety and expressiveness is hard to strike
properly, but that's why we've done all this work.
Peter
--
Peter Saint-Andre
https://andyet.com/
_______________________________________________
precis mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/precis
