Bryce: discussion? Cris: can we require the new specialization are randomized Antony:

if you want to have randomized stuff you need to change the initial of that. It’s an implementation question

Billi: Hash isn’t safe to use

Cris: and it’s slow

Bryce: there’s nothing that stops implementers from doing it

Titus: I don’t see discussion of whether it should be transparent

Antony: Transparent?

Titus: allow users to see the hash. For example tuple

Antony: was discussed in LWGI, if the types are same the hashes are the same

Titus: have we done enough thinking to identify scenarios, since we can’t add it after. We had a proposal in which we’ve decided on Umbviguasly transparent.

Ruslan: we do not just add heterogynous struct but also functions and array. Comparable is not cheap. It’s also

Antony: there are not very good outcomes. You can have inconsistently in the library. In addition, the Reflector discussion brought up transparent is not always a good idea, in case you want to keep it

Ruslan: I understand the consistency issue. If you need to compare string it will give a bad performance. We have measurements. Conversion operations take some cost.

Antony: we have to do some metafunction and decide if the hashing is expensive

Ruslan: we already know it’s expensive

Antony: depends on the hashing type you’re doing on the string

Ruslan: it’s not about the hash algorithm, the conversion costs

Billi: what will you make it transparent with? It’s unclear for this type. You probably want string view and const and what else? You want to not have

Titus: tuple of string string should be lookup with tuple string view

Ruslan: I’m saying we’ll be interested in transparent, not saying it’s easy

Antony: tuple from hash should give the same results, array of integers can also give the same results. (for transparent and not transparent)

Bryce: you think you can think about it today and come back tomorrow?

Billi: for the string cases the hashes already exist, so we can’t make them transparent.

Antony: users can always reuse the same cache, it doesn’t have to be transparent.

ну хотелось бы конечно для любых tuple alike во первых

Этот вопрос не ко мне, а к комитету. Я ценности именно в ISO тоже не вижу лично

посмотреть на другие модели развития языков без ISO, тип того

Видимо ценность видят большие корпорации.

в начале 90-х идея была в том, чтобы он стал площадкой для сбора экспертов, в первую очередь из корпораций еще читал недавно, что иногда для сертификации софта (вроде бы для автомобилей) его нужно писать на языке, у которого есть стандарт ISO

Сейчас-то зачем по таким правилам развиваться?

Даёшь ECMA C++!