[Haskell-cafe] Haskell 2014

[Tobias Dammers]

On Sun, Nov 30, 2014 at 04:25:36PM +0100, Nicola Gigante wrote:
> 
> Il giorno 30/nov/2014, alle ore 15:59, Roman Cheplyaka <roma at ro-che.info> ha scritto:
> 
> > The language («GHC Haskell») is evolving quite rapidly, it's just no-one
> > is really interested in maintaining the standard anymore.
> > 
> > I don't think it should disappoint you, unless you're a language
> > researcher or compiler writer.
> 
> Hi.
> 
> I’m a newcomer to the Haskell world, coming from C++ where the standard 
> and conformity to the standard is of great value.
> 
> Given the tendency of commercial implementors to deviate with custom
> and often bad-designed features, having an international standard that has to be
> followed by anyone is a great thing (and implementors are unfortunately
> very good at deviating anyway).
> 
> At first, the existence of the Haskell standard gave me a good impression.
> Haskell is not like other languages like python or Java that, at the end, have
> the One True Implementation. Haskell has born from the community, and
> there always have been a multiplicity of implementations. In this context,
> having a common standard to implement makes sense, to aid compatibility.
> 
> But Haskell is not like C++ neither. Haskell implementations are not driven
> by big corps, and features that deviates from the “standard” are not designed 
> and implemented by marketing departments, but they are instead often the
> implementation of new and innovative ideas from the research world. 
> 
> For this reason, it’s not so useful to crystallize the language to some-years-old
> standard when the compilers implementors, users and researchers are so
> good at evolving the language in a coherent way.
> 
> Here, I think, the point is the community: the language can continue to
> grow and evolve in the presence of multiple implementations by ensuring
> collaborations between the communities of the different compilers.
> If this continues to be done, I don’t think a formal standard, released 
> every x years, is needed.

Frankly, I believe that as it stands, there only really is one
industry-strength Haskell compiler - Hugs is dead, the others are either
research vehicles (interesting, ground-breaking, but far from being
useful alternatives to GHC for real-world applications IMO - I'd love to
be corrected on this one btw), or special-purpose tools (most notably
Haskell-to-JavaScript compilers like Fay) that don't even implement all
of Haskell 2010, let alone the recent additions found in GHC. The need
for a standard, therefor, isn't pressing enough at the moment, and the
kind of well thought-out and well documented development we're seeing in
GHC is enough to keep the language and its ecosystem moving at a high
quality. I can understand very well that people put their money, time,
and other resources, into actual features. That's fine, we currently
need those more than a standard.

If, at some point, an alternative compiler were to evolve (which,
personally, I would consider a positive thing: diversity and a bit of
healthy competition is good and can lead to great innovation boosts),
standardization would move up on the priority list, just like it did for
C++ when it became painfully obvious that GNU C++, clang-C++, Microsoft
C++, and other serious contenders, had produced an inconsistent and
incompatible mess between them that benefited nobody. We're *very* far
from this in the Haskell world, and from what I've seen in the
community, I don't think this is going to happen anytime soon, and if it
were, we'd see a new standard sooner rather than later.

(Note, btw., that one of the biggest sponsors behind GHC is Microsoft,
and many of the other contributions come from commercial entities -
Haskell consultancy firms, companies that use Haskell in their software
ecosystem, and even companies that are built around Haskell software.
Haskell has long ceased to be an ivory-tower research toy.)

My $0.02 anyway.