Re: [I18ndir] I-D on filesystem I18N

["Asmus Freytag (c)" <[e-mail]>]

On 7/7/2020 6:40 PM, Nico Williams wrote:
> On Tue, Jul 07, 2020 at 09:31:55PM -0400, John C Klensin wrote:
>> --On Tuesday, July 7, 2020 16:43 -0700 Asmus Freytag
>> <[e-mail]> wrote:
>>> However, early case-insensitive file systems did not preserve
>>> case. Not sure how rare this has become.
>> Well, Unix, and every Unix-derived system I know of (definitely
>> including Linux, FreeBSD, and NetBSD) are case-sensitive and
>> getting anywhere near their file names with Case Folding or even
>> lower casing will cause rather interesting problems.  One the
>> case-insensitive side, I've heard of a system called "Windows"
>> that usually preserves case but does not guarantee to do so and
>> many operations, in practice, don't.
> At least ZFS supports case-insensitivity as an option.  ZFS is available
> on just about every Unix and Unix-like OS today, and is quite popular.
>
> That said, I doubt anyone enables case-insisitivity in ZFS datasets on
> Unix unless they intend to share it with Windows clients and users
> expect case-insensitivity.  Indeed, that's the reason that ZFS has this
> option at all: Windows interoperability.
>
>> In addition, Asmus caught at least one case where your
>> terminology is inconsistent with that of Unicode.  Such
> Yes yes, I beg your forgiveness.  I wrote "normal form" instead of
> "normalization form".  A silly editing error.
Not silly, it may well be that "Normal Form" would have been the better 
term. Water under the bridge.

NFC/D (no-K) use what is called "canonical de/composition" in Unicode. 
Therefore, NFKx cannot be considered "canonical" or you will confuse 
anyone who knows Unicode usage.
>
>> inconsistencies (at least if not specifically identified and
>> justified) are an invitation to reader confusion or worse.  As
>> one more example, you should not be talking about case folding
>> and then lower casing.  They are different and have different
>> implications.  And, fwiw, most of my Turkish-speaking and
> Did you not read that text?  I only mentioned lower case in the context
> of explaining in introductory text what case-folding is.  And I did NOT
> say that they are the same thing.
Nico, please understand that we find your text otherwise quite readable 
and lucid - however, John's point may be that these concepts are subject 
to confusion (whether or not you share that) and that it would be useful 
to make extra-double sure that all readers get that these are different.

There was somewhere in the earlier e-mail discussion also the mention 
that is-lowercase and to-lowercase (however you name these two concepts) 
are crucially different in that there's an expectation that is-uppercase 
= ! is-lowercase (ignoring is-titlecase for the argument), but that 
to-lowercase and to-uppercase are operations that may be inverses of 
each other. German to-lowercase(to-uppercase(ß)) != ß (but "ss").

It would be good to lay the foundation by reminding everyone of these 
subtleties.
>
>> writing colleagues would disagree that U+0131 "could" or should
>> be considered equivalent to U+0069.  In my experience, that
> The idea was to describe how an admin might square that circle in a
> case-insensitive filesystem, and then I did not endorse it.  The idea
> was to describe how custom case-folding tailorings (FYI, that's the term
> Unicode uses, 'tailoring') might come about.
>
> There will be time to address all the nits.  The big idea is more
> interesting.

The danger of reviewing the test while reading along from the start. 
However, the only way I know how to review anything that's not in a 
change-tracked editable format.

A./

>
> Nico