I am watching this discussion from the sidelines, and I am little involved in the standardization. But I think judging from the outside, Vinnie has a point. Some recent additions to the standard made questionable design choices, which if a library had been implemented and widely used prior to standardization like in Boost, design choices may have been made differently. Some examples: - The distinction between views and containers is academic, and after standardizing ranges, the distinction had to be watered down to make ranges practical. - std::format is a parallel development to ranges, outputting into a sink instead of being itself conceptually a range of characters. - The distinction between weak and strong ordering is academic, I never found a good use case in our code where making the distinction would be beneficial. - Both are not an enum, and not being able to switch over them is a pain. I am all for implementing and trying things out first before we commit to a design. Boost was a very good staging area for that. Arno -- Dr. Arno Schödl CTO We are looking for C++ Developers: https://www.think-cell.com/career/dev think-cell Software GmbH (https://www.think-cell.com) Leipziger Str. 51, 10117 Berlin, Germany Main phone +49 30 6664731-0 | US toll-free +1 800 891 8091 Amtsgericht Berlin-Charlottenburg HRB 180042 Directors: Alexander von Fritsch, Christoph Hobo Please refer to our privacy policy (https://www.think-cell.com/privacy) on how we protect your personal data. schoedl@think-cell.com | +49 30 6664731-0> On 10. May 2024, at 00:25, Vinnie Falco via Boost <boost@lists.boost.org> wrote:
On Thu, May 9, 2024 at 3:07 PM Vinnie Falco <vinnie.falco@gmail.com> wrote:
...
I pressed send by accident... continuing...
When people vote on proposals, it is based on the honor system whether or not they have read the paper, if they are qualified, if they are knowledgeable about the domain, and if they are voting in good faith. WG21 has to simply trust that someone who votes yes is not doing so because the author voted yes or will vote yes to their own paper. WG21 has to simply trust that when someone votes no, they are doing so because they believe it is the best technical decision and not because they simply don't like the author's politics. Or criminal record. Or because they or their company intends to introduce a new competing paper in the future.
WG21 leaves questions of conflicts of interest, political horse trading, non-technical voting up to the honor system. No one who votes is vetted for their talent, industry experience, and so on. Peter Dimov, probably the smartest guy on the planet, had his metaprogramming library turned away by people with far less talent. The consequence is we do not have Peter's library, but instead the promise of a better metaprogramming standard library component that has yet to be written.
I do not have this luxury when I write libraries. I can't go up to say, a large corporation, and convince them that I'm a really great guy who should just be trusted. People have to opt-in to my library, unlike the standard where after a relatively small group of people vote, all vendors who produce standard libraries are compelled to add it. And every C++ user who uses that standard library now has that pushed on them. No one "forces" you to use standard library components. But the appearance of a feature or particular API in the standard library creates enormous resistance to alternatives, because there is value in having a normalized API which is bundled with the compiler.
:LEWG is now notorious for having people write "direct-to-standard" proposals. That is, people find it far easier to just write a paper and socially engineer their way through WG21's tyranny of democracy than to invest the blood, sweat, and tears of writing a popular library. It has literally been said "going through Boost is more work than going through LEWG." In other words the bar for technical excellence in LEWG is lower than it is for Boost. I, for one, am glad that I am not someone who "have [sic] attended rather more of its meetings than the two people who make just slightly questionable claims about what WG21's library standardization process is fit for" because if I was, then I would be responsible for that lowering of the bar.
As I believe that complaining without offering solutions is equivalent to "whining" I propose a simple solution. Eliminate LEWG, have library-only components go straight to LWG like how it used to be, ensure a process where people voting on papers are actual subject matter experts and not patsies or confederates, and require that these library-only components already have some level of adoption and represent the state of the art. Note that under this scheme, we would already have the Networking TS in the standard. And the standard library would be able to connect to the Internet.
Thanks
_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost