The Ignorability of Attributes in C++

a year ago

Attributes in C++ were initially designed to be ignorable, which has led to them being unsuitable for solving many problems.
The introduction of attributes like [[override]] and [[final]] as contextual keywords instead of attributes has led to inconsistency and complexity in syntax.
The decision to make attributes ignorable has harmed language evolution, leading to worse decisions and more discussions about self-imposed problems.
Examples like [[no_unique_address]] show how ignorable attributes can lead to ABI breaks and compatibility issues.
Current language features, such as trivial relocation in C++26, suffer from being implemented as contextual keywords instead of attributes, leading to more complexity and less expressiveness.
Future language features could benefit from non-ignorable attributes, but the current model restricts their usefulness.
The article argues for abandoning the idea of ignorable attributes to stop harming language evolution and to simplify future feature additions.

Hasty Briefsbeta