D is our last hope

[Guillaume Piolat]

On Tuesday, 19 December 2023 at 14:57:00 UTC, Hors wrote:
> On Tuesday, 19 December 2023 at 14:28:15 UTC, Luna wrote:
>> Additionally, I feel all these discussions about "D is dead, 
>> what should we do" is a waste of air. If you want D to be more 
>> active, contribute to the ecosystem, make PRs to projects so 
>> that we don't have so many one-man-operations going on. Not to 
>> mention an increase in financial support to D developers 
>> working on the ecosystem, such as WebFreak making code-d, dub 
>> developers, etc, would also increase activity within the D 
>> ecosystem.
>
> "contribute to the ecosystem" is really unclear. What we can 
> actually contribute, new libraries? Changes to language? If 
> it's changes to language, then what we can add (or rework) to 
> the language?

I feel tempted to repost this:
https://dl.acm.org/doi/10.1145/2509136.2509515

In absence of new evidence, we should assume nothing makes as 
much difference as open-source libraries. The same paper says 
that the language features matter less so. So it's not very 
moving when people talk about languages or stdlib features on 
these forums, understandably the focus as a language community, 
but still.

So this library-first strategy is exactly what Grim is/was going 
on about with DlangUI, as it's a library packed with features 
that can make a large difference for D. The examples are very 
impressive!

How do we leave the state where it's a ton of work for a few 
people?
I think I may have a solution there for large frameworks:
https://dplug.org/tutorials/Dplug%20Tutorials%2016%20-%20Dplug%20architecture%20and%20D%20ecosystem.pdf

People that like to build :) can design more of their 
sub-libraries to be reusable for others, hence sharing a 
maintenance load more. It's a bit hindered by the various 
fragmentation factors ("is there a runtime? a GC? a libc?").

The problem is that traditionally library design in D follows the 
STL bottom-up STL Stepanov-style, but we should really make 
domain-specific libraries that follows the good practices of 
popular C libraries. That also, would be evidence-based.