Re: [boost] Re: string algorithms review

From: Peter Dimov (pdimov_at_[hidden])
Date: 2003-11-03 07:44:14

David Abrahams wrote:
> "Peter Dimov" <pdimov_at_[hidden]> writes:
>
>> Depends on where we want Boost to go. Do we want just a repository
>> of fairly useful libraries (for some broad definition of "useful")?
>> Or do we want Boost to be a "sandbox library working group", where
>> library authors go to get a feel of how the committee operates, to
>> establish the infamous "existing practice" that is often needed for
>> a standard library proposal, and so on?
>>
>> This question has never been given a clear answer; I suspect that
>> the reason is that most Boost members want to have both. But "having
>> both" will break sooner or later, as Boost grows.
>
> Why do you think so?

Because, as more and more divergent libraries are accepted, there will be
more and more pressure (from both sides) to move Boost in either direction,
and because in case of conflicting priorities, there is no guidance how to
resolve the conflict and obtain that accept/reject vote.

>> It may already have broken but we can't see it yet.
>>
>> My personal opinion is that we should move away from the "anything
>> goes as long as it's useful" model and move even closer to a
>> "sandbox LWG" model; I'd go even further and suggest that we should
>> have an "official" paper submission process and an "official"
>> agenda. In a "sandbox LWG" reviewing a library without documentation
>> is simply unacceptable, although reviewing a documentation without
>> library is not!
>
> While I agree that documentation should be a prerequisite for a
> review, I think you may underestimate the value to the standardization
> process of having a fairly broad scope of libraries in Boost, even if
> some may never be suitable for the standard.  Since one major aim of
> the next standard is to support library development, it's important to
> be able to cite Boost's existing practice where functionality is
> needed.  Justifiably or not, IMO other libraries don't get the same
> respect.  Also, our vision today of what's suitable for
> standardization is considerably evolved from what it was two years
> ago.  If we fail to consider some libraries on the basis of their
> "projected standardizability", we may find ourselves missing one or
> more libraries that really should be in the WP two years from now.

I don't underestimate the value (either sign) to the standardization process
of a diverse Boost... but what I'm talking about here is the value of the
formal review from the library author's point of view, if the author plans
to submit a standard library proposal at a later date. In the long run, it
is probably better to endure a "harsh" Boost review, however painful that
might be, and then breeze through the LWG review (relatively speaking)
than... the alternative. ;-)