Date: Sat, 1 Aug 2020 00:28:45 -0400
On 7/27/20 8:28 AM, Alisdair Meredith via SG16 wrote:
> I have started updating P2139 following the telecon last week, and my current
> draft can be found at:
> https://isocpp.org/files/papers/D2139R3.html
>
> I’d appreciate anyone who attended that can confirm whether I correctly reflect
> the position of the group.
Looks good over all. A few comments.
D.20:
* With regard to:
"to avoid premature breakage of clients of the UTF-8/UTF-32 parts
of this feature ..."
I think the concern was less about breaking clients and more the
removal of a feature for which the standard does not currently
provide a replacement.
D.21:
* Just a point of clarification in case it isn't well known or
understood. The non-deprecated std::codecvt facets from <locale>
cannot be directly used with std::wstring_convert and
std::wbuffer_convert because they have protected destructors. A
user provided facet with a public destructor is required (such a
class can inherit from a std::codecvt specialization;
cppreference.com provides examples).
D.22:
* With regard to:
"The specific facets in D.22 stand in the way of putting a
minimally useful replacement into the standard, by sitting on the
good names but with poor semantics"
I don't think this captures the reuse concern that was discussed in
the meeting. These facets don't prevent providing a minimally
useful replacement; they squat on the specializations that could
potentially be wanted for adding support for conversions between,
for example, UTF-16 and the narrow encoding as opposed to between
UTF-16 and UTF-8. But as discussed in the meeting, the motivation
for pursuing such reuse of these specializations is very low.
* I think the reason that the zombie clause doesn't apply is because
the zombie clause applies to names and these deprecations apply to
specific specializations of a non-deprecated name. Presumably, if
the deprecated specializations are removed, an implementation could
choose to re-purpose them.
D.23:
* With regard to:
"SG16 is not persuaded that this is actually a text issue ..."
I think the concern wasn't whether u8path is a text issue, but
rather that, since a suitable replacement exists, whether to
preserve or remove the legacy interface is more of a LEWG concern
than an SG16 one. Per P1253 <https://wg21.link/p1253>, we do
consider handling of filenames to fall under our purview.
Tom.
>
> I will add poll numbers when the meeting minutes are available.
Minutes are now available at
https://github.com/sg16-unicode/sg16-meetings#july-22nd-2020.
<https://github.com/sg16-unicode/sg16-meetings#july-22nd-2020>
Tom.
>
> Thanks
> AlisdairM
> I have started updating P2139 following the telecon last week, and my current
> draft can be found at:
> https://isocpp.org/files/papers/D2139R3.html
>
> I’d appreciate anyone who attended that can confirm whether I correctly reflect
> the position of the group.
Looks good over all. A few comments.
D.20:
* With regard to:
"to avoid premature breakage of clients of the UTF-8/UTF-32 parts
of this feature ..."
I think the concern was less about breaking clients and more the
removal of a feature for which the standard does not currently
provide a replacement.
D.21:
* Just a point of clarification in case it isn't well known or
understood. The non-deprecated std::codecvt facets from <locale>
cannot be directly used with std::wstring_convert and
std::wbuffer_convert because they have protected destructors. A
user provided facet with a public destructor is required (such a
class can inherit from a std::codecvt specialization;
cppreference.com provides examples).
D.22:
* With regard to:
"The specific facets in D.22 stand in the way of putting a
minimally useful replacement into the standard, by sitting on the
good names but with poor semantics"
I don't think this captures the reuse concern that was discussed in
the meeting. These facets don't prevent providing a minimally
useful replacement; they squat on the specializations that could
potentially be wanted for adding support for conversions between,
for example, UTF-16 and the narrow encoding as opposed to between
UTF-16 and UTF-8. But as discussed in the meeting, the motivation
for pursuing such reuse of these specializations is very low.
* I think the reason that the zombie clause doesn't apply is because
the zombie clause applies to names and these deprecations apply to
specific specializations of a non-deprecated name. Presumably, if
the deprecated specializations are removed, an implementation could
choose to re-purpose them.
D.23:
* With regard to:
"SG16 is not persuaded that this is actually a text issue ..."
I think the concern wasn't whether u8path is a text issue, but
rather that, since a suitable replacement exists, whether to
preserve or remove the legacy interface is more of a LEWG concern
than an SG16 one. Per P1253 <https://wg21.link/p1253>, we do
consider handling of filenames to fall under our purview.
Tom.
>
> I will add poll numbers when the meeting minutes are available.
Minutes are now available at
https://github.com/sg16-unicode/sg16-meetings#july-22nd-2020.
<https://github.com/sg16-unicode/sg16-meetings#july-22nd-2020>
Tom.
>
> Thanks
> AlisdairM
Received on 2020-07-31 23:32:07