Re: Matching and comparison

[James Seng <jseng@pobox.org.sg>]

One suggestion which we can think along is we do not do any case folding at
all, including English. this would break the current English system to certain
extent since RFC1035 Section 2.3.3 states that case insensitivity comparsion
should be done.

On the other hand, the same section also state that "loss of case sensitive
data must be minimized" so things are may not be too bad...

Microsoft ID say that all US-ASCII charset are lowered case at the resolver
end before it is send out where others remains unintact. This is good guide
but may be insufficient for our purposes.

-James Seng

"Martin J. Duerst" wrote:
> Turkish has a dotted i and a dotless I. For Turkish, converting
> a I into an i is like converting an o to an u for English.
> Most probably, we cannot change that DNS treats i and I the
> same already. But we want iDNS not to do that for Turkish.
> It is okay (well, even desired) to treat the two dotted i's
> ('i' and its uppercase) and the two dottless I's (I and its
> lowercase) the same. Obviously, the server doesn't know what
> language the user speaks, and the server could know that a
> label is Turkish, but maybe we don't want to go there, so
> the best think to do, among many non-perfect alternatives,
> may be to say, as the Microsoft I-D did, that the client/
> resolver lowercases things. On a Turkish system, I would be
> lowercased to a dotless i, and an I with a dot would be
> lowercased to i. It is true that this distroys the uniqueness
> of mappings between e.g. print and machines, but the alternatives
> I have seen are all worse (throwing dottless and dotted i
> into one pot for the Turkish, and so on).