Proposed resolution:

Add a new paragraph to the specification of affine_on in [exec.affine.on] providing a specification for the default implementation:

-6- Let sndr and env be subexpressions such that Sndr is decltype((sndr)). If sender-for<Sndr, affine_on_t> is false, then the expression affine_on.transform_sender(sndr, env) is ill-formed; otherwise, it is equivalent to:

  
auto [_, sch, child] = sndr;
return transform_sender(
  query-with-default(get_domain, sch, default_domain()),
  continues_on(std::move(child), std::move(sch)));

except that sch is only evaluated once.

Proposed resolution:

Add a new paragraph to the specification of affine_on in [exec.affine.on] providing a specification for the default implementation:

-6- Let sndr and env be subexpressions such that Sndr is decltype((sndr)). If sender-for<Sndr, affine_on_t> is false, then the expression affine_on.transform_sender(sndr, env) is ill-formed; otherwise, it is equivalent to:

  
auto [_, sch, child] = sndr;
return transform_sender(
  query-with-default(get_domain, sch, default_domain()),
  continues_on(std::move(child), std::move(sch)));

except that sch is only evaluated once.

[ 2025-10-17; Reflector poll. ]

Set priority to 2 after reflector poll.

The wording of affine_on doesnt have a specification for the default implementation. For other algorithms the default implementation is specified.

The intention for affine_on was to all optimisation/customisation in a way reducing the necessary scheduling: if the implementation can determine if a sender completed already on the correct execution agent it should be allowed to avoid scheduling. A specification should provide enough lattitude to allow that.