Re: Boost vs Std library features

From: Niall Douglas (s_sourceforge_at_[hidden])
Date: 2023-11-29 14:26:10

On 29/11/2023 13:18, Peter Dimov via Boost wrote:
> Andrey Semashev wrote:
>> I'm not a committee member, just an observer from the outside, but it
>> looks to me that people that aren't familiar with the proposal should
>> abstain from the discussion. And if those people constitute the majority
>> of the room then either the proposal discussion should be rescheduled to
>> a later meeting or the room should be dismissed to do their homework.
> 
> That's not going to work. You go to Kona and the committee is like mate,
> we haven't bothered to read your paper, why don't you wait six months
> and come to Tokyo instead.

The chair advertises in advance what papers will be seen at a meeting, 
those who are in the room at the time are on person's honour they've 
done the due diligence involved to speak usefully about a paper.

What happens in practice is a lot of people in the room when the paper 
changes load up the paper on their laptops, scan it very quickly, and 
then make somewhat informed comments with varying signal to noise ratios.

For my papers which went to normative wording, I decided to drop all 
arguments about design, and just list out a condensed design summary as 
LEWG agreed (in some case years ago), the change revision log since 
then, and the current proposed normative wording.

This prevents rapid page scanning in the way people are used to, though 
the condensed design summary is intended to let them get up to speed 
within three pages of reading intensively instead of forty pages with 
far more words. I therefore get constant comments and complaints about 
me having "unhelpfully" removed the forty plus pages of arguments about 
design. Anybody who makes such a comment is effectively saying "no I 
didn't read your paper before right just now", though they probably 
don't realise that they're admitting that in public.

In the end, you can't make people in the room study a paper beforehand, 
or if they didn't, to not make unproductive comments. The current 
process is fundamentally broken in my opinion. WG21 should adopt a 
better one, ideally one not based on library proposal papers consisting 
of dozens to hundreds of pages of proposed normative wording.

We have a far more terse library specification language available to us 
than ISO normative wording. It's called "reference implementation".

Niall