On Fri, 22 May 2020 at 15:44, Peter Dimov via Boost
Mathias Gaunard wrote:
Nothing prevents you from using a pool allocator with a deque.
And nothing prevents me from using the proposed ring_queue instead. Do you have a better point than "you can achieve the same effect in a more complicated and stupider way"?
I don't think combining two well-established generic tools in a way they are meant to be is either of "complicated" or "stupid". Of course you can always roll out something special for every use case. But that comes with a lot of overhead of maintaining it compatible with other components, flexible, efficient and bug-free. Less code is better than more. Ultimately the fact that uses are not too niche or that there is significant value on top of existing solutions are criteria for illegibility for Boost. Not that those criteria are very heavily enforced, as we have a lot of libraries that only target domain specialists, so it's up to how strongly people feel about things.