On Fri, Apr 18, 2025 at 1:56 PM Ivan Matek via Boost <boost@lists.boost.org> wrote:
On Fri, Apr 18, 2025 at 4:47 PM Peter Dimov via Boost < boost@lists.boost.org> wrote:
On 18 Apr 2025 04:51, Peter Dimov via Boost wrote:
The domain-specific expertise should come from the reviewers; the review manager should (minimally) just be qualified enough to
evaluate
Andrey Semashev wrote: the
reviews.
Reviewers are allowed to be less familiar with the domain, and in fact may not be familiar at all. There were plenty examples of such, including myself in the recent Boost.Hash2 review. The review manager's job is to weigh the quality and relevance of the reviews, and to be able to do that the review manager has to be at least as qualified as the reviewers, very preferably more.
For Hash2, we got a review from one of the authors of Blake2 (if we knew that in advance we'd certainly have included Blake2 in the initial submission.)
I don't think it would have been realistic to expect or demand equal or higher level of expertise from the review manager in this case.
I understand the difficulties of finding a review manager, but in this discussion I am mostly agreeing with Andrey point. To use your example: there was no guarantee that author of Blake2 would submit a review. And in my opinion at least one person who is judging the library needs to be a domain expert. Now I see two options:
1. keep the current requirement that manager is a domain expert 2. drop the requirement that manager needs to be an expert if we are confident expert(s) will submit reviews
This way we have at least one domain expert involved with approval. Issue here is that this complicates the process, by requiring reviewers to agree beforehand to submit a review, compared to current one where it is much easier for reviewers to submit a review.
Just to be clear I do understand how difficult is to find anybody to do unpaid work, not to mention highly skilled domain experts to do unpaid work, but I believe that beside author at least one person involved in review needs to be an expert in the domain, and that general C++ expert knowledge is not enough in general case(counter example would be variant2 that is "pure" C++ library without need for reviewers or manager to be experts in unicode or networking or coroutines or ...).
A problem with that thought I see is that we don't have any guarantee of domain experts doing reviews. And, honestly, I think being a domain expert gets in the way of being an impartial review manager. And I consider being an impartial manager is rather important. As domain experts often inject their own opinions (likes and dislikes) into the decision. Which can be counterproductive when it comes to evaluating what we hope are forward thinking modern library submissions. -- -- René Ferdinand Rivera Morell -- Don't Assume Anything -- No Supongas Nada -- Robot Dreams - http://robot-dreams.net