On Sunday, 3 April 2016 at 09:12:05 UTC, Joakim wrote:
I though Laeeth had a good suggestion on how to market D a
couple months ago, as the current front-page pitch may be too
general for some chunk of readers:
"A set of 'channels' for different use cases might be helpful.
Eg bioinformatics, numerical computing, web, etc. Both for
tutorials and setting out the advantages."
I'd organize it by adding a usage page to dlang.org with a list
of popular channels like that, with a paragraph of info for
each and links to the wiki with more info about use in that
field. The usage page would have some links from the front page
pitch.
As such, we need to collect info on how you all are using D
now. If you are using D in some field like that, please
describe what you're doing and we'll add it to the website.
Not exactly what I had in mind longer term (combination of shiny
appealing thing that gets the idea across quickly would
compliment grittier unpolished stories), but here is a start:
https://wiki.dlang.org/User_narratives_on_switching_to_D
When you're at the periphery for the time being, you have to give
some thought to what your advantages are, what problems other
people have, and how you can get them to see how you can solve
it. But of course first of all you may need them to help
perceive and articulate that they have a problem - amidst the fog
of life, that's not always as clear as it sounds when described
after the fact by a writer like Michael Lewis (Big Short, Liar's
Poker).
And I think given where D is its much more likely to be
recognised as the right tool for many jobs by hearing how someone
else - that's a real person not a symbol - had a problem that is
kind of just like yours but was able to make use of this workshop
(as Walter describes it) to ease their pain and do things that
never were quite easy enough to tackle before.
You have to make the best of what you got. People are very
helpful here - its not just John Colvin rewriting someone's
ported Python code, and Adam Ruppe solving someone else's obscure
linking problem (by doing so he is responsible for what is likely
D's first adoption by a hedge fund - Andy Smith's old shop) as
well as 10 questions a day on irc. On the other hand, people
arent shy about describing what they don't like, and that's
ultimately a positive because it means you have a community that
is more concerned about quality than being agreeable for the sake
of it. And, as Andrei says, be honest till it hurts, is a great
way to move towards excellence. From a marketing point of view
then, it's better to make a clean breast also of the areas where
D isn't where we would want it to be - yet. By doing so it's
then much easier to put things into proper context and shape the
narrative. When I started with the language it really wasn't
clear to me that it was suitable for using in production, and if
you read social media then you might still come away with that
impression. Clarity may be the best answer to fear,
uncertainty, and doubt.
And delivering on what you say you will is a great way to build
trust. If you come to D thinking your python code when ported
directly will automatically be 200x faster (I exaggerate for
effect) then when it's only a few times as fast you will be
disappointed. So then its better to explain that, and why, and
that for processing large amounts of data the GC isn't great, but
at the same time you really don't need to make such a fuss about
it - D is not Java, and its easy to avoid the major infelicities
and here are some examples of what you can do.
Anyway, we aren't going to draw users by means of shiny marketing
primarily. The latter does matter because if you are in an
organisation and take a social, political, and commercial disk by
doing something different, then its much easier for you if others
who check it out take a quick superficial look (because
heuristics in the face of attention and time starvation), but you
most of all want to appeal to the guys who want to believe anyway
and just need to have some help in understanding that this can be
the case.
People in similar lines of work often have similar problems. So
then its very interesting if you are doing finance or
econometrics work to be shot instantly not just to Bachmeier's
work (different things, but opening up R libraries is huge), but
also to hear in practical concrete terms what he uses it for and
what the benefit is.
Similarly if rapid prototyping is important - which ought to be
the case for many - then Walter's Worp experience is very
powerful. Yes,it's the language creator who is also a very good,
performance oriented programmer (and would he have achieved a
comparable speedup had he been working in Perl ? ;) but a story
doesn't need to be objectively indubitable even to a sceptical
observer to get the point across. You will find an account of
that if you watch lots of YouTube videos or happen to stumble
across the Wired article, but in general the people in
organisations with the power to make decisions don't - except at
the very top - have so much time to do this.
If I recall correctly plasticity is mentioned on the front page,
but it's not made super clear what that means and the benefits
arent made vivid if you arent a!ready aware of them. I don't
think there is a link to a summary of Walter's mention of it in
relation to the practical experience of developing Worp or
Andrei's recapitulation of the experience - and no nice little
summary that gets me interested enough to watch the whole talk.
Similarly there is no one place go learn about how people are
using it in bioinformatics - there was even an academic paper on
the benefits, but I don't think you would find that easily from
the front page. (The audience for the channels isnt limited to
only people in that sector, as I guess if you process a lot of
string data, probably you take peek at what the bioinformatics
guys are doing). And how many prominent clicks does it take to
go from the front page to an understanding of who Sociomantic
are, what they do, at what scale, and how much they save vs their
competitors by using D rather than the alternatives. Or again,
AdRoll is a python shop - they have given talks about their use
of python at scale. But their data scientists use D. That's
funny - Guido says python is fast enough. Why do they use a
language with a smaller ecosystem and worse tooling than the
obvious alternatives ? What has been their experience in
practice, warts and all ? Maybe they don't want to talk about
it, but has someone approached them and asked if they would like
to ? I would do it myself - I did write some wiki content, but
I just don't have the time, and will try to help longer term in
other ways. But it wouldn't take much work to make some progress
- so much low hanging fruit, and initially its perhaps more about
creativity and coming up with a plan than it is sheer grind work.
Beginnings are often modest, and one doesn't need to drink the
ocean in one gulp - just working away at it will work wonders, I
think.
Laeeth.
