Re: Boost, not LEWG

From: Niall Douglas (s_sourceforge_at_[hidden])
Date: 2020-07-01 09:31:09

On 01/07/2020 05:44, JeanHeyd Meneide via Boost wrote:
> So I'm going to chime in here, in aggregate. I will try to focus on
> answering the question of "What is Boost Good For?" from the
> perspective of a user, and as a user having participated in reviews.
> It's a large brain dump, so there's probably typos and other bad
> things.

Firstly, this is a great reply. Erudite, useful, positive,
actions-based. Anybody who hasn't read all of it ought to.

What is especially noticeable is the focus on the end user experience
rather than constant giving out about other people and orgs not doing
things right in some Boost people's opinion.

> Not having the library -- and encouraging the text layer not to come
> back as part of the next resubmission of the library -- is a huge blow
> to "this really addresses one of the serious contemporary C++
> problems". Maybe Zach will resubmit that layer, but we missed out on
> the sweet spot of addressing a need that neither Standard C++ nor most
> shops handle nicely. I do hope the library comes back!

I could not agree more with this point! Whilst yes, Unicode is the pain
point in the ecosystem, and yes, more standards targeted libraries ought
to go through Boost before LEWG, and yes, the most efficient use of
Zach's limited spare time is to concentrate on what is (relatively
speaking) easiest, it is actually replacing **std::string** which is the
golden calf needing slaughtering here.

I wouldn't wish achieving that on my worst enemy. It would be awful. But
it would also come with outsize improvements across the ecosystem in all
sorts of non-obvious ways. Like Outcome had. Improving strings and text
in C++ would be incredibly hard, but if successful, incredibly beneficial.

> Boost may need to actually reach out to people and very much handhold
> them through the process of library submission, review, and
> improvement to convince a few key cases that Boost is still extremely
> relevant to modern needs. Maybe that will bring in even more folk;
> encourage them to submit here, once they see that Boost libraries are
> once more the place to go to solve real problems, with a tangible
> feedback loop. Right now it's hard to gauge the effect of a Boost
> library, since Boost is doing very little to distinguish itself from
> the Standard (again, why go through Boost if you can do GitHub + LEWGI
> + LEWG + LWG instead?). Boost should be the cutting edge again. This
> does not mean the quality has to go down: but there should be active
> encouragement of exploration of new topics OR addressing people's
> needs. Boost.Text, Boost.Unicode, maybe even Boost.Audio, and more...

I don't personally think that would work. The cost benefit to getting
into Boost just isn't there. I did it once. I'll never do it again.

What I do think would work is if Boost v2 becomes a peer anointed
collection of Boost-endorsed third party libraries bundled into a
distribution i.e. getting into Boost means that your library ships with
every Linux and Windows distro thenceforth, a common distribution channel.

This means that "getting into Boost" shifts from the author doing the
driving into the Boost community choosing a set of rules by which
popular C++ libraries are selected for candidacy (e.g. >= 5 years
maintained, permissive licence, STL naming conventions, widely portable
etc), and then persuading the author to undertake any changes which a
peer review decides upon.

That effectively turns Boost into a highly selective package manager. I
find that just fine.

(Note that current Boost v1 ought to remain as-is, package-manager Boost
would be a separate v2)

> Basically, mailing lists are wild-wild wests of formatting and
> quality. A platform that got rid of that etiquette and just let me
> conform to an established format (GitHub Issues, etc.) greatly lowers
> the barrier to being able to speak up.

I find the mailing list stuff greatly overdone. "Young people" have no
problem with mailing lists if there is a motivation to join. Back when I
was applying myself at Boost, we had a constant incoming stream of new
blood, because I was going out there and setting up a pipeline to bring
young people into Boost. Some even tried their hand at getting a library
into Boost. Only one succeeded, but all of them figured out mailing
lists just fine, and they joining up here caused other young people to
also sign up. But since I moved on, that pipeline has dried up, so
mailing list traffic has dropped, vitality has ebbed etc etc.

Participation and relevance doesn't "just happen". It occurs when an
investment in making it happen occurs over many years. WG21 grows every
year because there is a bunch of people investing in making it grow
every year. Boost has not grown every year recently because the people
who were investing in its growth a few years ago have moved on, so right
now there is a lull. I'd like to hope that instead of complaining about
others over which Boost folk have no control and indeed just look bitter
by the constant complaining, some here might consider stepping up and
investing effort outside the ivory tower, create vitality, restore
relevance, by *engaging* with the ecosystem and its current needs
instead of complaining about it.

Niall