On Mon, 9 Mar 2026 at 18:39, Jonathan Wakely <cxx_at_[hidden]> wrote:

>
>
> On Mon, 9 Mar 2026 at 18:01, Halalaluyafail3 <luigighiron_at_[hidden]>
> wrote:
>
>> On Mon, Mar 9, 2026 at 5:08 AM Jan Schultke <janschultke_at_[hidden]>
>> wrote:
>> >
>> > I think there's another useful guarantee that could be provided, but
>> I'm not sure whether it should be done as part of this proposal: In any
>> encoding, it looks like alphabetic characters always appear in their
>> lexicographical order, even if they aren't always contiguous.
>> >
>> > Consequently, (a <=> b) is meaningful and useful between two characters
>> if you know that a and b are letters.
>>
>> Letters of the same case. To make it somewhat useful between different
>> cases it
>> would require guaranteeing 'Z'<'a' or 'z'<'A' being true. Only somewhat
>> related,
>> but the C++ standard does not currently have a definition of letter like
>> C does.
>> Therefore it is unclear whether or not this is valid:
>>
>> int main(){int _\u0393;/*reserved?*/}
>>
>
> There is no reason for the implementation to reserve such names. It would
> be a no-op to change [lex.name] to say uppercase letter from the basic
> character set.
>

We do define 'letter', in the library's [character.seq.general] subclause.
But it is defined in terms of uppercase letters which still isn't defined.