The forked elephant in the room

Steven Schveighoffer schveiguy at gmail.com
Fri Jan 19 22:34:55 UTC 2024


On Friday, 19 January 2024 at 20:22:07 UTC, H. S. Teoh wrote:
> D has always been closer to the closed development model where 
> Walter has final say over what goes in and what doesn't. Not 
> quite as closed as Lua (devs don't even grant access to the 
> code repo, you only get the code snapshot at release time, they 
> may or may not listen to community feedback and are not 
> obligated to explain why), but still pretty near the closed end 
> of the spectrum than your average open source project where the 
> community's voice plays a bigger role.

This is a bit far. Development of the compiler and the libraries 
are very open. In as far as Phobos (and to a great extent 
druntime) is concerned, Walter is not even involved. 3rd party 
projects that are essential to D's ecosystem are not actually 
even owned by DLF, but are still considered critical 
infrastructure.

Adding *features to the language* is a different story. It is 
very easy to "just add this new thing" without thinking about the 
far-reaching consequences. That will lead to a disaster IMO. I'm 
actually glad that DIP1036 (not the most latest thing, which was 
a modified version) did not just make it in on the first round, 
and we worked through the debates and came to 1036e. I've had 
other features that I invented that got added which are less than 
ideal and hard to correct (inout for instance).

I think D has done pretty well with Walter at the helm, and I'm 
not convinced the alternative would have been better. I don't 
want to make excuses, because I know what it's like to be on the 
receiving end of dismissal (it's one of the reasons I really 
don't try to make any sweeping changes to Phobos any more), but I 
think the current leadership situation is fixable, and we are 
better off trying to fix it than seceding.

> There are pros and cons whichever way you take.  I have my 
> opinion on where on the spectrum things are more ideal, but 
> it's not possible to know for sure without actually doing it.  
> It's not an easy issue.

I would expect most open source to be designed and modified with 
one person or a small team at the top dictating what is OK and 
what is not, with many others who are trusted contributors. D is 
not any different.

-Steve


More information about the Digitalmars-d mailing list