Subject: Re: [boost] Boost.Fit review Mars 3-20 result
From: Robert Ramey (ramey_at_[hidden])
Date: 2016-05-18 14:46:00
> Robert, Rob, you are right that I could have rejected the library but
> I've preferred to accept it subject to a new review. The main reason is
> that I believe that this library would be useful for the C++ Boost
> community. It's not ready yet, but the features the library proposes are
That's the way I read it.
> Rob, the conditions will not be for conditional acceptance, but for
> having a new review. Sorry if this was not clear. with my sentence
>> Conditional acceptance (a new review is needed)
OK - I think that was widely misunderstood.
> Unfortunately I've no time to do the report of the modifications that
> will be needed before the new review. The deadline for Oulu C++ standard
> committee is end of this month and I have some papers to finish.
> Please, be patient,
No problem - I understand that it takes time to good job reviewing all
the reviews. This is especially true when you don't really want to
dismiss the submission outright but hope to see it transformed in some way.
May main concern was it seemed accepted into boost subject to some
conditions which were expected to be met - if not proforma. It seems
that I got this wrong. I've objected in the past to the
characterization of "conditionally accepted" in this context. Though
I'm sympathetic to the idea that submission to boost should not be a
gratuitously unpleasant process, I think attempting to "soften the blow"
can lead to more frustration than less.
I would also like to see us encourage pre-reviews on the incubator. I
know this idea has not been successful, but I still think it's a good
idea. I think it can be hugely helpful and motivating to library
developers, improve quality of submissions and raise acceptance rate for
library submissions. It can also wake up a library author to the fact
that he's not really ready to submit - at much smaller cost in emotional
frustration. For an example of how this can work, see the review in the
safe numerics library. This one review made me re-think fundamental
aspects of the submission in a way that otherwise wouldn't have
happened. As I responded to the review and subsequent comments, I did a
couple of things:
a) I made the library simpler. It move toward a clearer and more
unified concept as opposed to a disjoint bag of tricks.
b) I had to make a better case about why I thought such a library was
needed and worthwhile. The resulted in creating a bunch of tutorial
examples illustrating the case where I thought such a library was
necessary. Of course more examples made me think about the focus and
feature set that was actually required.
After this exchange - I worked another year on the library.
If this library were to come up for review, one review is already done,
in the queue and available for reference by other reviewers and/or
review manager. I believe that the effort invested in this pre-review
process will more than recovered in saving time in any final review and
avoiding the dreaded "conditionally accepted" which is like your future
ex-girlfriend telling you "I like you as friend".
Also note that reviews posted in the incubator would be available to
future users of the library and would be easily available to explain the
rationale for all the decisions made - and all the features left out
and all the arguments about this. It serves as an expanded FAQ.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk