Date: Tue, 7 Mar 2023 10:57:56 -0500
SG16 will hold a telecon on Wednesday, March 8th, at 19:30 UTC (timezone
conversion
<https://www.timeanddate.com/worldclock/converter.html?iso=20230308T193000&p1=1440&p2=tz_pst&p3=tz_mst&p4=tz_cst&p5=tz_est&p6=tz_cet>).
*This message will also serve as your friendly reminder that this
meeting is taking place tomorrow. **I'm sorry for, again, publishing an
agenda so very late. *
The agenda follows.
* P2773R0: Considerations for Unicode algorithms.
<https://wg21.link/p2773r0>
Corentin provided an overview of this paper during the 2023-02-22 SG16
telecon
<https://github.com/sg16-unicode/sg16-meetings#february-22nd-2023>. For
this telecon, I'd like to focus on:
1. Understanding, from an interface perspective, what additional
requirements are needed to implement tailored algorithms, and
2. Evaluating consensus for recommendations and conclusions presented
in the paper.
A list of candidate polls follows. Please feel free to comment or
suggest additional ones.
* Unicode algorithms should be provided as views even if eager
implementations that perform better are possible and/or provided.
* Unicode algorithms provided as views should operate solely on
sequences of code points (not code units; a decoding view can be
provided as input).
* Transformational Unicode algorithms (normalization and casing) are
not applicable to encodings that cannot encode the entirety of the
Unicode character set (without resorting to some form of
transliteration).
* char32_t should be used as the sole type for Unicode code points.
* Character data provided in storage other than char8_t, char16_t, and
char32_t must have an explicitly associated character encoding (an
encoding such as the literal or locale encoding is not implicitly
assumed for char or wchar_t).
* Character properties that are not immutable (e.g., that won't change
between Unicode versions, even for unassigned characters) should be
concealed from ABI and ODR concerns (thus preventing their use in
constant evaluation).
* It is not necessary that efficient implementations of non-tailored
Unicode algorithms be implementable using ICU.
* It is necessary that standardized interfaces for Unicode algorithms
be implementable such that they are competitive (within an order of
magnitude) with ICU performance using idiomatic ICU interfaces.
Tom.
conversion
<https://www.timeanddate.com/worldclock/converter.html?iso=20230308T193000&p1=1440&p2=tz_pst&p3=tz_mst&p4=tz_cst&p5=tz_est&p6=tz_cet>).
*This message will also serve as your friendly reminder that this
meeting is taking place tomorrow. **I'm sorry for, again, publishing an
agenda so very late. *
The agenda follows.
* P2773R0: Considerations for Unicode algorithms.
<https://wg21.link/p2773r0>
Corentin provided an overview of this paper during the 2023-02-22 SG16
telecon
<https://github.com/sg16-unicode/sg16-meetings#february-22nd-2023>. For
this telecon, I'd like to focus on:
1. Understanding, from an interface perspective, what additional
requirements are needed to implement tailored algorithms, and
2. Evaluating consensus for recommendations and conclusions presented
in the paper.
A list of candidate polls follows. Please feel free to comment or
suggest additional ones.
* Unicode algorithms should be provided as views even if eager
implementations that perform better are possible and/or provided.
* Unicode algorithms provided as views should operate solely on
sequences of code points (not code units; a decoding view can be
provided as input).
* Transformational Unicode algorithms (normalization and casing) are
not applicable to encodings that cannot encode the entirety of the
Unicode character set (without resorting to some form of
transliteration).
* char32_t should be used as the sole type for Unicode code points.
* Character data provided in storage other than char8_t, char16_t, and
char32_t must have an explicitly associated character encoding (an
encoding such as the literal or locale encoding is not implicitly
assumed for char or wchar_t).
* Character properties that are not immutable (e.g., that won't change
between Unicode versions, even for unassigned characters) should be
concealed from ABI and ODR concerns (thus preventing their use in
constant evaluation).
* It is not necessary that efficient implementations of non-tailored
Unicode algorithms be implementable using ICU.
* It is necessary that standardized interfaces for Unicode algorithms
be implementable such that they are competitive (within an order of
magnitude) with ICU performance using idiomatic ICU interfaces.
Tom.
Received on 2023-03-07 15:57:58