Boost logo

Boost :

From: David Abrahams (dave_at_[hidden])
Date: 2007-06-07 00:50:15


on Wed Jun 06 2007, "Peter Dimov" <pdimov-AT-mmltd.net> wrote:

> Eric Niebler wrote:
>> Peter Dimov wrote:
>
>>> Why do we need a review manager at all?
>>
>>
>> Primarily to avoid any questions or doubts about whether a library
>> should be accepted or not. The review manager supposedly takes
>> everybody's feedback into account, but makes the ultimate yes/no
>> decision, and is even free to buck popular opinion.
>
> Do so many of our reviews end in such a non-conclusive manner as to require
> a decision from a review manager?

IMO yes, enough do. Also, a cursory reading of the votes does not
always give enough information to understand whether the library
should be accepted. Some votes are made with less consideration than
others. IMO it's important that somebody write up the decision and
the rationale for the decision, so even if the consensus is apparent,
the review manager has an important role to play.

>> Another key job of the review manager is to collect all the
>> feedback and present a TODO list to the author of the new Boost
>> library, but that's secondary.
>
> I think that the author should be quite capable of doing that
> him/herself.

That would be nice, but IMO new authors will often be inclined to
discount or discard things that would otherwise become important
requirements.

> This also applies to most of the other review manager duties. The
> rest can be handled by documentation. The final yes/no decision to
> accept can be done by the moderators.

That's definitely not a responsibility I want.

> Going by the checklist one by one:
>
> * Checks the submission to make sure it really is complete enough to
> warrant formal review.
>
> Not really required. The reviewers will be quick to point that out.

Sure, but meanwhile we've wasted everyone's time and occupied a
valuable review slot.

> * Asks the review wizard for permission to extend the review schedule if it
> appears that too few reviews will be submitted during the review period.
>
> The review can take as long as necessary to gather a sufficient number of
> reviews. There is no need for a deadline. If we decide to keep the current
> scheme, the submitter can ask for the extension.

A deadline is an incentive for people who want their vote counted to
actually get on with doing a review.

> * Decides if there is consensus to accept the library, and if there
> are any conditions attached.
>
> It is the responsibility of the submitter to prepare a summary of the
> reviews linking to them and to work with the reviewers to address their
> concerns. The summary is posted to the list and the moderators decide
> whether to accept the library.

Puts too much burden on the moderators, IMO. How do we know that all
the important concerns have been addressed? The review manager has to
follow the review process in detail and make sure nothing important is
being stepped over.

> * Posts a notice of the review results on the regular boost mailing list,
> the boost-users mailing list, and the boost-announce mailing list.
>
> The moderators do that.

Another responsibility I don't want.

IMO, aside from not having enough volunteers at the moment, a
condition I believe we can fix in any number of ways, the review
manager role, and reviews in general, are a part of the Boost process
that I think really works and doesn't need fixing.

FWIW, I have to admit that, although I think it's good to re-examine
everything we do from time to time, I am a bit uncomfortable with
re-examining everything we do from the ground up all at once, which
seems to be the direction in which we're now headed.

-- 
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com

Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk