On Wednesday, 7 February 2018 at 21:02:11 UTC, data pulverizer
wrote:
On Tuesday, 30 January 2018 at 20:45:44 UTC, Andrei
Alexandrescu wrote:
https://www.quora.com/Why-hasnt-D-started-to-replace-C++
Andrei
The Betamax Problem
When you introduce something new, how do you know that it is
going to be compelling enough for people to move from whatever
it is they are doing and use your new thing?
It is not an easy question to answer but in the realm of
programming languages it's a very tough question, because
people are going to have to learn a whole new language, and its
going to come with costs and potentially unquantifiable risks
for any company that attempts to shift to that language. So
whatever it is you are offering has to be tremendously
compelling compared to what is already there.
A great deal of confusion in the world arises from failing to
make distinctions between things that appear to be the same but
really aren't when you look closely.
Also, in about 1870 odd there was a revolution in economic
thought that took place more or less simultaneously in Vienna,
Lausanne and Cambridge. The Marginal Revolution had yet to be
fully digested in the way people think about social phenomena.
My director of studies at Cambridge, Lord Eatwell, Labour
spokesman in the House of Lords, was known for his devotion to
the work of Pierro Sraffa, a man known principally for a rather
hostile book review of a book by Hayek and a rather slim book of
his own, Production of Commodities by Means of Commodities, a
book that tried to draw insights about the economy from a model
with two goods, corn and gold.
And I think considering firms as homogeneous, with the same
cultural values and facing the same situation will be about as
insightful as I think Sraffa's work ended up being - not very.
Life is risk. It's risky to get out of bed in the morning, but
it's also risky not to get out of bed. And it's true that an
agent acting on behalf of someone else - ie a manager who has no
stake in the business - will often think about things first in
terms of not taking a decision that might lead him to be blamed.
But most firms are not large enterprises, and in the US small and
medium sized firms over time create more than 100% of job growth,
last I checked. And a manager who is also at least to some
extent a principal ie an owner in the business knows that to be
conservative in a time of change is not necessarily prudent, and
it may well also not be the profit maximising choice.
As someone who is both a manager and a part owner I disagree that
a new technology choice needs to be overwhelmingly compelling to
be considered. And I don't get paid to make decisions about
things that are easily quantifiable - what for you need me for if
the numbers are straightforward?
The reality is that firms are very different, in a dynamic
industry even within the same sector they are different. And at
any one time there are a bunch of people close to trying D or
more. You don't need to persuade everyone to grow. You just
need to persuade a few more people to tip over the margin. And
there are often plenty of safe ways to take risks. You just need
to make sure you have a plan B. Listen to Manu's talk for a real
example of what I mean. And note that he said Finns are very
conservative.
An important question is what problem set does D solve? It's
very hard to sell a language to industry without convincingly
answering that question. If you are selling them a 'better'
language - that's a tougher sell. If you are selling a solution
to a particular problem set - you stand better a chance.
But really who is selling D to anyone? We are very far from that
stage right now. Did someone sell D to Microsoft COM team,
Remedy or to Weka? Nope. People who had earned the authority to
decide became aware of the language end decided to use it. And
they did so because for them it solved their particular problems
better then anything else they could think of.
For a manager to consider D as the successor to C++, it doesn't
just have to be a better language design than C++, it has to
have the best language design of any compiled language and
demonstrate the best performance.
Why? Best in what way? Best for whom and for what kind of
problems?
I completely disagree with that. It needs just to be better in
the situation then the conceivable alternatives. And situations
and challenges are really quite different between firms.
Is the former really true?
Are various language features that have been inherited from
C++/Java the best way forward? For instance does D have the
best approach to object oriented programming, or templates? Or
any important set of features you care to mention? Are there
things that C++ does better than D? How straightforward is it
to get great performance from D? Is how do you 'tune' your D
code for high performance obvious or well documented?
If the answers to any of the above questions is a negative for
D, that's a serious problem if what you want to do is replace
C++, because C++ is already a solid well know language and the
competition from new programming languages is extremely tough,
and because its only going to get easier to create programming
languages this competition will get tougher.
To finish the Betamax story, when CD came along, people dropped
cassette tapes like hot potatoes and DVDs killed VHS stone
dead. Does D represent a similar leap from C++?
I'm not saying that D is Betamax. I'm just giving food for
thought.
You can turn your scepticism on anything, and the value of the
exercise really depends on the results. But doubt on its own
isn't necessarily a good guide to the future in my experience.
In particular with D because the benefit doesn't I think come
from any single feature but from the totality of the gestalt of
what it's like to work in D compared to other languages.
I think that if you want to understand the adoption of new
technologies then Clayton Christensen's Innovators Dilemma has
considerable insight. Up to a point Crossing the Chasm and
companion books too. I think by Moore.
