On Sat, Jun 20, 2020 at 4:35 PM Zach Laine <whatwasthataddress@gmail.com> wrote:
On Sat, Jun 20, 2020 at 3:10 AM Andrey Semashev via Boost <boost@lists.boost.org> wrote:
On this note I'm going to ask something probably obvious, but still:
1. Why do you have to embed ICU code or data in Boost.Text? Why can't you link with it?
Because that implies a 20% slowdown (more, in cases where you're not using the exact pointer types used by ICU), and of course a dependency on ICU.
To be clear, I think this is the least bad option on the table at the moment. I'm not happy with losing that extra perf, but I also think it's more important to get the interfaces into the hands of users, as this library is explicitly targeting standardization. I honestly had not considered the fact that including ICU code under the Unicode license in a Boost distribution of course requires all users to review the Unicode license in addition to the Boost one. I agree that this is not a reasonable way forward. Zach