Re: Review Manager Volunteers

From: Robert Ramey (ramey_at_[hidden])
Date: 2025-04-22 19:57:32

On 4/17/25 6:51 PM, Peter Dimov via Boost wrote:

>>> I agree, and.. I've wondered about that requirement. Is it really
>>> needed? If an established Boost developer has a good amount of
>>> professional experience they are likely to be able to tackle most
>>> programming domains at the level needed for managing a review. Should
>>> we reconsider that requirement? For example I would consider myself to
>>> know enough to manage the Bloom review. But...
>>
>> Given that the review manager is the one solely responsible for deciding
>> whether a library deserves acceptance, it would be strange not to require a
>> certain, fairly high level of expertise.
> 
> The domain-specific expertise should come from the reviewers; the review
> manager should (minimally) just be qualified enough to evaluate the reviews.
> 
> General C++ expertise is more important for the review manager because
> he is supposed to help the submitter prepare the library so that it's
> suitable for Boost.

I think it's pretty hard to set a specific policy here.

Take a library like serialization.  Most of us would be comfortable 
being review manager as we know what the domain is about and something 
about the features which different approaches.

Take a library like safe_numerics.  I think this would require a review 
manager with above average knowledge of numerical analysis.

So I think the current process for selecting review managers is fine. 
Of course there will be issues from time to time.

1) Sometimes the review manager way underestimates the amount of time it 
will take to do the review.  So the result is delayed.  In at least one 
case I know of the review manager flaked out entirely.  I'm sure there 
are more such chases.

2) It's hard to find a qualified reviewer for some libraries.  No easy fix.

The only thing that occurs to me that in our 
communications/announcements we increase the level of recognition for 
review manager - more or less to  the level of that of author.  This 
would be in the hope of making RM to be sufficiently prestigious that he 
might be motivated to note the fact that he was RM on his resume and/or 
other credentials.

Personal experience note: I'm the author of the boost safe numerics 
library.  I developed it incrementally over many years on my own until I 
felt it was ready to submit to boost.  It had all the boost 
requirements: documentation, tests, etc. etc.  I submitted it and it was 
accepted into Boost with conditions. I believe that Andrey Semashev was 
the review manager.  I should know this for sure off the top of my head. 
  The fact that I don't illustrates my point above.

I was ecstatic to get this news and went right to work making 
modifications to full fill the conditions and other issues Andre had 
flagged on his own.  In doing this task, I realized that the library was 
much, much lower quality than I had thought.  In cleaning up these 
"nits" I found lots and lots of other issues : concepts not agreeing 
with code, undetected errors, etc, etc.   The same thing happened with 
the serialization library ~24 years ago.  At that time I had an excuse: 
I was still learning C++ (I'm forgetting it now).  But this time ...  I 
was honestly appalled and disheartened with my own work.  With Andres 
help, I managed to fix everything.  But I've learned so much in the 
meantime, that I'm tempted to make another pass over the library to make 
it more formally correct and eliminate unnecessary requirements.  From 
feedback I get from users, and some statistics, I believe that the 
library has pretty good up take.  I don't think this be so much the case 
without Andre's collaboration.  This may not be true for other authors 
and their libraries, but it's a fact that it's true for mine.

So maybe we should highlight/promote more of the contributions of 
reviewers and review manager in our documentation and promotional 
material (announcements of library acceptance, etc.). (do we have 
promotional material?) In fact we should spend a little more effort in 
promoting boost, libraries, authors and reviewers.  Did you know that 
the boost serialization library is used in the software that runs 
"icecube" - huge project in antarctica which I believe is used to study 
neutrinos or something like that.  I only found out about this by 
accident (no not that kind of accident). It's a huge motivator for me to 
be associated with a project like that - albeit peripherally.

Robert Ramey