Right now, what's the most important for the success and adoption of D?

[Brad Roberts]

Daniel Keep wrote:
> I can't really disagree with what anyone else has said so far, so I'll
> throw in something new:
> 
> A new D compiler that is 100% independent of the DigitalMars code base.
>  Doesn't have to be efficient, doesn't have to have awesome code
> generation, but it needs to exist and it needs to be conformant.
> 
> And it needs to be written in D itself. :)
> 
> Why?  Take a look at how changes get added to Python.  Someone comes
> along, and thinks "hey, X would be a useful feature!"  They introduce
> the feature to the newsgroup, write up a Python Enhancement Proposal,
> and discuss it.  In D's case, we stop about here (albeit without any
> kind of formal proposal system).  With Python, they then typically go
> off and write a patch to the interpreter to implement the change so that
> people can experiment with it to see if it's useful or not.  Having an
> easy to modify D compiler written in D would make this much easier, and
> allow proposed features to actually be prototyped instead of those
> long-winded entirely hypothetical threads on the NG where people argue
> endlessly about a feature they can't actually test.
> 
> There's also the benefit that implementing the language spec will force
> the language to be completely defined; it'll highlight holes, and ensure
> the spec makes sense.
> 
> Also, we can finally say to people: "yes, D does have a compiler
> implemented in D itself."  Not that it really matters, but perception
> does. :)
> 
> Anyway, just thought I'd add something instead of just parroting.  Now
> if you'll excuse me, I'm off to get some biscuits...
> 
> 	-- Daniel

While a D compiler in D itself might be interesting in some ways, it'd 
be, well, highly non-trivial to produce.  Instead of investing man years 
of effort duplicating the front end, optimizer, and code generator that 
lives in dmd, how about finding a project that is in need of developers 
and work on it instead?  It's not like there aren't tons or that we're 
overflowing with manpower to spend on redundancies.

For those interested in experimenting with adding features to D can pick 
up gdc and play around _right now_.  For an awful large subset of 
changes, they would be isolated to the front end parts exclusively and 
wouldn't need to get terribly involved with the gcc layers.  Doing that 
has the additional advantage of having the potential to actually produce 
a diff that Walter can use as a basis for actually adopting the change.

Anyway, just my opinions.

Later,
Brad