Re: Boost vs Std library features

From: Niall Douglas (s_sourceforge_at_[hidden])
Date: 2023-11-29 11:55:04

On 29/11/2023 04:24, Vinnie Falco via Boost wrote:
> On Tue, Nov 28, 2023 at 6:49 PM Jeff Garland via Boost <
> boost_at_[hidden]> wrote:
> 
>> I assume you mean when LEWG basically didn't exist and it was *just LWG*
>>
> 
> This structure (with "just LWG") was vastly superior to the current model
> (LEWG + LWG) in terms of delivering the best value for C++ users. Because
> when it was just LWG, the group was composed also of great library
> designers in addition to wordsmiths. By creating the new group LEWG, the
> great library designers are stuck in LWG where they cannot change anything
> given to them by LEWG, and meanwhile LEWG is populated by people who just
> want to "get their library into std" for various personal reasons.
> 
> Furthermore having "just LWG" created a natural bottleneck: not every
> library could go in, so decisions weighing the benefits and comparing what
> brings the best value for the limited committee resources were made more
> conscientiously.

I would go further than that Vinnie - back when all of WG21 sat in the 
same room, library and language worked together a lot better.

The current situation at WG21 - as proved in multiple votes over 
multiple years - is that Library have diametrically opposite opinions on 
what is important in many important topics than Language.

This manifests in EWG refusing to consider language changes requested by 
Library, thus forcing Library to either make do or invent some hidious 
metaprogramming hack which adds seconds per compilation unit because 
nobody over in Language will entertain that Library might have valid 
concerns. You then get complaints from some in EWG that Library is 
"doing all the wrong things", and "only if it were run more like EWG".

Obviously returning everybody to a single room isn't practical. But I 
also think using the same processes for Library as for Language is a bad 
move. Library isn't like Language, it should have its own unique and 
quite different processes in my opinion.

In particular, I think Library should gain the power to change the 
language where it suits Library, and EWG removed from that part 
entirely. That will be seen as incenduary I am sure, however I think we 
would have long ago solved library ABI stability mitigations and 
lightweight exception handling if Library were allowed to decide some 
parts of language design.

Niall