Re: [Boost] [Describe] Summary Review Summary

From: Edward Diener (eldiener_at_[hidden])
Date: 2021-03-18 18:04:58

On 3/18/2021 12:07 PM, Niall Douglas via Boost wrote:
> On 17/03/2021 18:02, Emil Dotchevski via Boost wrote:
> 
>> If your goal is standardization, convincing the users is utterly irrelevant
>> to success. Worse, it is a lose-lose proposition, you might get one and a
>> half stars on GitHub which doesn't look too good. I remember Niall giving
>> (good) advice that if the goal is standardization, it is best to not bother
>> with a Boost review, either: it adds a lot more work that is irrelevant to
>> achieving your goal, plus you risk rejection which doesn't look too good.
> 
> That's not _quite_ what I advised, though it is close.
> 
> My advice was, and always has been, that the most valuable aspect of
> Boost _to the library_ and its author is the peer review. A high quality
> review is quite literally priceless - it cannot be bought for money.
> It's why I get so annoyed when some proposers only care about getting
> into Boost at all costs, that the review is only a hurdle which can be
> beaten down using groupies, which feels to me an enormous wasted
> opportunity to make the best C++ libraries possible.
> 
> In this aspect, Boost has been very good for me. Both of the libraries I
> presented to Boost - AFIO and Outcome - both were completely
> reimplemented and completely redesigned from the feedback received here.
> I personally believe that the reason why WG21 has to date liked my
> libraries so much is precisely because of those Boost reviews. So thank
> you Boost!
> 
> 
> What you may be remembering me saying Emil is that once you get the
> review from Boost, the incentive to finish the library and get it into
> Boost is low if your goal is standardisation. You can skip rewriting all
> that documentation and tutorials and examples and having to deal with
> real end users for the next two decades if you go straight from Boost
> review to WG21 standards, skipping finishing your library sufficient for
> Boost. Given the plethora of C++ package managers today including
> github, Boost as a distribution vehicle is nothing like as important as
> it once was, so all in all, the value added from shipping in Boost is a
> fraction of what it was fifteen or twenty years ago.
> 
> I skipped Boost and went straight to WG21 with LLFIO, and I feel very
> guilty about it. I did do the honourable thing for Outcome however. And
> I'll never, ever, get another library into Boost again unless someone is
> paying me to do it. Once is enough.

Was it really that bad, Niall ! You have already said that the review 
was very helpful. Most good C++ programmers probably view getting a 
library into Boost as an indication of their ability and also a great 
experience in hearing what their peers think. On the practical side 
getting a library into the C++ standard must be far more difficult than 
getting a library into Boost, so I do not think it is very practical for 
a programmer to think that some library he has developed is going to be 
accepted as part of the C++ standard, but it may be good enough to get 
into Boost. The focus here is very different from the C++ standards 
committee as far as libraries are concerned, thank goodness. So despite 
your viewpoint I would be willing to bet that most C++ programmers find
Boost much less onerous when it comes to library approval than the C++ 
standards committee. This is not meant to be a criticism of the C++ 
standards committee in any way. They are the custodians, if you will, of 
a computer language and their viewpoint as to what constitutes that
language and its official software has to be stringent as far as 
large-scale usefulness is concerned.