stability

[Derek Parnell]

On Sat, 23 Feb 2008 21:49:42 -0800, Walter Bright wrote:

> Bill Baxter wrote:
>> I think Walter is doing a pretty good job of fixing bugs with D1, but 
>> still there are a few instances where he has seemed a little too eager 
>> to classify a fix as an enhancement, apparently so as to avoid having to 
>> port it to D1.
> 
> That isn't true at all. "Porting" changes to D1 is trivial, since the 
> code is about 98% common. The problem is there is often not a clear 
> distinction between a bug and an enhancement, and D1 cannot be a stable 
> language if it keeps getting enhancements.

A bug is where the current behaviour is contrary to the specification. 
An enhancement is proposed behaviour that is not documented in the
specification.

The problem we (including Walter) have is that D does not have a
specification and therefore there is no method to determine if a given
behaviour is by design or a mistake. All we have is "common sense", "that's
how C++/Java/C#/XXX does it", "it feels right", etc ...

So I would imagine that before another D compiler is written, a
specification is created that forms the benchmark to measure what is a
conforming D compiler. 

-- 
Derek Parnell
Melbourne, Australia
skype: derek.j.parnell