From: Asmus Freytag (asmusf@ix.netcom.com)
Date: Wed Aug 22 2007 - 00:29:40 CDT
On 8/21/2007 8:49 PM, Philippe Verdy wrote:
>> part de Asmus Freytag
>> Envoyé : mardi 21 août 2007 09:31
>> À : verdy_p@wanadoo.fr
>> Cc : unicode@unicode.org
>> Objet : Re: New Corrigendum to The Unicode Standard
>>
>> On 8/18/2007 7:27 PM, Philippe Verdy wrote:
>>
>>> Can we say somewhere that Unicode 5.0 without the corrigendum is no
>>> more compliant,
>>>
>>>
>> Who's the 'we' here?
>>
>
> Don't know, anyone that considers this being important.
>
The only body that rules on versions of the Unicode Standard is the
Unicode Consortium. And the settled position of that body is that once a
version of a Unicode Standard is published it remains valid and *unchanged*.
>
>> The Unicode Standard (and website) make very clear that a corrigendum
>> does not actually modify a version.
>>
>
> It does so in this corrigendum, because the BiDi UAX is fully part of the standard, making Unicode 5.0.0 without corrigendum incompatible with Unicode 5.0.0 with this corrigendum 6 for the BiDi algorithm.
>
No. Unicode 5.0.0 is unchanged. It produces results that may not be
satisfactory in terms of mirroring these characters, but it remains
valid and well-specified. Implementations that claim to be conformant to
Unicode 5.0.0 would be expected to produce a certain set of results.
Only if you also claim conformance to the corrigendum would you get a
different set of results, one that most users would consider more correct.
> That's why I suggested an easy way to designate versions with or without corrigenda. OK "d" is taken by drafts, lets' use "c" for corrigendum level, this does not change the problem.
>
> So the Unicode 5.0.0 book without corrigendum would be 5.0.0c0, and with corrigendum 6, it would be 5.0.0c6
>
Corrigenda are exceptional by nature, and it's worth calling explicit
attention to them. What is so bad about saying "Unicode 5.0.0 with
corrigendum 6 applied". It is much clearer than saying 5.0.0c6.
One problem with your notation is that corrigenda are not successive.
There are some versions of Unicode to which more than one corrigendum
might be applied. But it is not required that implementers apply any of
them in sequence.
Your discussion below seems to not have taken that fact into account.
Corrigendum 5 is not higher than Corrigendum 4. Either, none, or both
might have been applied to certain versions of Unicode, Unicode 3.2.0 in
this case. So now you have four branches instead of one.
Version 3.0.0 could have any combination of corrigenda 1,2,3 and 5 (but
not 4) applied. And so on. This information is easily gleaned from
http://www.unicode.org/versions/corrigenda.html and your attempt of
creating an numbering scheme and graphs is not only redundant, but may
actually be harmful because it seems to easily mislead the reader into
assuming a linear nature for corrigenda.
A./
PS: skipping from here since it appears to be based on a wrong premise.
The information presented in the graph below is also incorrect.
> As opposed to version major and minor numbers, the upward compatibility is
> not guaranteed between two successive corrigenda, in fact each corrigenda is
> creating its own branch, and the lowest corrigendum level is already out of
> the trunk in the compatibility tree, only the highest corrigendum level is
> in the trunk, versions in the branches are not mutually compatible, but the
> nearest from the trunk has the lowest cvompatibilitydifferences with it, so:
>
> (Latest)
> ||
> || 5.0.0c0 (TUS 5.0.0 book, without the corrigenda)
> || |
> || 5.0.0c1
> || |
> || 5.0.0c2
> || |
> || 5.0.0c3
> || |
> || 5.0.0c4
> || |
> || 5.0.0c5
> || |
> 5.0.0c6 |
> || |
> |+--------+
> ||
> 4.1.0c0 (has no corrigendum)
> ||
> || 4.0.1c0 (TUS 4.0.1, without the corrigenda)
> || |
> || 4.0.1c1
> || |
> || 4.0.1c2
> || |
> || 4.0.1c3
> || |
> || 4.0.1c4
> || |
> |+--------+
> ||
> 4.0.1c5
> ||
> 4.0.0c0 (TUS 4.0.0 book, has no corrigendum)
>
>
> And so on...
>
>
>
>
>
>
>
>
This archive was generated by hypermail 2.1.5 : Wed Aug 22 2007 - 00:32:56 CDT