DConf 2013 Closing Keynote: Quo Vadis by Andrei Alexandrescu

[Joakim]

On Thursday, 27 June 2013 at 03:20:37 UTC, Mathias Lang wrote:
> I've read (almost), everything, so I hope I won't miss a point 
> here:
> a) I've heard about MSVC, Red Hat, Qt, Linux and so on. From my
> understanding, none of the projects mentionned have gone from 
> free (as in
> free beer) to hybrid/closed. And I'm not currently able to 
> think of one
> successful, widespread project that did.
Then you are not paying attention.  As I've already noted several 
times, Visual Studio, which is the way most use MSVC, has both 
paid and free versions.  Red Hat contains binary blobs and 
possibly other non-OSS software and charges companies for 
consulting and support.  Qt is an "open core" project that is 
dual-licensed under both OSS and commercial licenses, the latter 
of which you pay for.  Linux contains binary blobs in the vast 
majority of installs and most people running it paid for it.

If your implied point is that the original authors aren't the 
ones taking the project hybrid or paid, it depends on the 
license.  Sometimes it is those owning the original copyright, as 
it had to be in the Qt, MySQL, and other dual-licensing cases, 
other times it isn't.

> b) Thinking that being free (as a beer and/or as freedom), 
> hybrid, closed
> source of whatever is a single critera of success seems 
> foolish. I'm not
> asking for a complete comparison (I think my mailbox won't 
> stand it ;-) ),
> but please stop comparing a free operating software with a paid 
> compiler,
> and assume the former have more users than the later because 
> it's free (and
> vice-versa). In addition, I don't see the logic behind 
> comparing something
> born in the 90s with something from the 2000s. Remember the 
> Dot-com bubble ?
Obviously nothing is a "single criteria of success," as has been 
stated already.  In complex social fields like business or 
technology ventures, where there are many confounding factors, 
judgement and interpretation are everything.

By your rationale, we might as well do _anything_ because how 
could we possibly know that C++ wasn't immensely successful only 
because Bjarne Stroustrup is a Dane?  Obviously none of this 
discussion matters, as D has very little Danish involvement and 
therefore can never be as popular. ;)

You have to have the insight to be able to weigh all these 
competing factors and while I agree that most cannot, those who 
are successful do.

> d) People pay for something they need. They don't adopt 
> something because
> they can pay for it. That's why paid compiler must follow 
> language
> promotion, not the other way around.
These assertions are somewhat meaningless.  Those who value 
performance will pay for the optimized version of the dmd 
compiler that I've proposed.  Those who don't will use the 
slower, pure-OSS version.  There is no reason for a paid compiler 
to only follow promotion, both must be done at the same time.

In any case, I've lost interest in this debate.  I've made my 
case, those involved with the D compiler can decide if this would 
be a worthwhile direction.  From their silence so far, I can only 
assume that they are not interested in rousing the ire of the 
freetards and will simply maintain the status quo of keeping all 
source public.

This will lead to D's growth being slowed, compared to the 
alternative of providing a paid compiler also.  That's their 
choice to make.

If somebody stumbles across this thread later, perhaps they will 
close up optimization patches to ldc and sell a paid version.  
Given that those behind dmd have not expressed any interest in a 
paid version, maybe these ldc vendors will not involve them with 
the money or feature decisions of their paid ldc.  It would be 
likely that this paid compiler becomes the dominant one and the 
original dmd project is forgotten.

If you don't choose the best approach, a hybrid model, you leave 
it open for somebody else to do it and take the project in a 
different direction.