A suggestion to D community

[Sean Kelly]

Sergey Gromov wrote:
> Bill Baxter <wbaxter at gmail.com> wrote:
>> On Tue, Sep 16, 2008 at 11:07 PM, Yonggang Luo <luoyonggang at gmail.com> wrote:
>>> Rayne Wrote:
>>>
>>>> http://www.dsource.org/projects/tangobos
>>>>
>>>> Tada!
>>>  I know of that, but it's don't solve the fundamental problem. And it work separately.
>>> I means that the Tango group and Phobos group and other group can do some communication. Let's Tango and Phobos using the same runtime. And we create seperate project for Operating System Binding, And seperate project for language (such as c) binding. So every one just need to considerate the staff he focus on.
>>> Nowadays, Phobos and Tango have it's own binding. And phobos can't not working with tango officially.
>>>
>>> I means that the official group can do these things. And we create a project to collect all of these project to reduce the working of the D Programmer. They just need to download a single package and support tango and phobos offically. And these package also containg the runtime library for DMD, GDC, LLVMDC and so on. And these project will link the official runtime library of DMD,GDC,LLVMDC.
>>>
>>> If  we do this, DMD, GDC, LLVMDC will create it's own runtime library. Because every compiler's runtime library is not the same. and A group do the effort for common part of the runtime library. And tango group just focus on the tango library, don't need to consider the dmd,gdc, llvmdc and so on. Like this, will reduce the repeatedly working.
>>>
>>> But this need all of us to do the effort.
>> This sounds like another flavor of the "mythical man-month" fallacy.
>> The idea that if you just throw more people at a project it will
>> succeed.  First of all, you need people with very special skills to do
>> the kinds of things you're asking people to do.  Second even if you
>> round those people up, they will all have day jobs, and other
>> obligations, so you're going to get maybe a few hours a week out of
>> most of them.  One hour a week from 40 people is nothing like 40 hrs a
>> week from one person.  Third if they really believed what you are
>> suggesting is a good idea, they would probably already be doing it.
>> Fourth I'm not even really sure what is is you're asking people to do,
>> and chances are that I'm not the only one who is thinking your plan is
>> rather vague.  What exactly does it mean to you to create a new
>> "runtime" library?  You mean like a competitor to Phobos and Tango?
>> Creating a third runtime is the solution to too many runtimes?   I
>> just don't get it.
> 
> I think that a huge step towards unification would be dividing Phobos 
> into Phobos and Core, where Core is an absolute minimum required to 
> support all built-in language constructs.  I.e. GC, implementation of 
> arrays, implementation of all built-in operators on all built-in types.  

Tango is already designed this way.  The runtime component is completely 
isolated from the rest of the library.

> So that Phobos is built on top of the Core.  So that Tango team could 
> say that here's Tango, and here's an alternative Core which is 
> compatible with the original but has these fixes and those absolutely 
> necessary extensions which are proposed for adoption by the reference 
> Core.  So that the Core is standard and libraries are pluggable.

It's more likely that this common core will come from Tango, since I've 
already done all the necessary work for this.  Stay tuned.


Sean