[erlang-questions] OT: Programming Language Selection as a Business Strategy
Mon Apr 16 06:19:44 CEST 2007
Ulf thanks for the long detailed reply.
Ulf Wiger wrote:
> It's probably more accurate to say that Erlang was developed in an Ericsson
> laboratory in response to a perceived technology need. Bjarne Däcker's
> thesis gives some good background on this:
I've grabbed a copy and sent it to the printer to read.
> It also mentions that Erlang was banned in 1998, due to perceived
> business needs.
> (For the newcomers, this ban preceded, and in part justified, the launch of
> Open Source Erlang.)
> I agree with Thomas that essays like "beating the averages" give an
> angle on this question. Large companies will also worry about the cost of
> diversity and lock-in effects. This is most likely much less of a
> problem in
> small companies.
> For large companies, aspects that will have measurable consequences include:
> - Cost of training support staff in several different technologies,
> since different
> products are developed differently
> - Difficulty in aligning user interfaces and documentation, since
> products do
> not share a common implementation base.
> - Re-training costs when moving developers from one project to another
> - Difficulty in moving product development responsibility from one design
> center to another.
> - Difficulty in finding third-party suppliers and consultants for obscure
> - Risk that a niche technology folds during the life of the product, forcing
> an expensive redesign.
Valid points that explain some of the success of java and C/C++ over
some "better" alternative and is exemptified be this quote
"We were after the C++ programmers. We managed to drag a lot of them
about halfway to Lisp."
- Guy Steele, co-author of the Java spec From
> Some would also want to include the need to use mainstream languages
> since this is what job applicants are looking for, but I think this is
> a phallacy, even though in a narrow perspective, the argument does seem
> to hold some merit.
Agreed, I know a few developers, myself included, who only use
"mainstream" languages as a last resort. One is running a start-up with
a few friends developing a HR webapp in php, another working for a
university but run link alert/alarm or similarly named company on the
side with his wife (who is the main developer) using perl, another is
working for an ISP using some scripting language (slips my memory at the
moment). Don't see much evidence of mainstream languages in this bunch.
> Now for the problem: If you're looking to invest, say, $100 million in a
> product development project, wouldn't you want to try to eliminate the
> above challenges, if possible? And knowing how many other factors
> influence the success of a large development project (large = more than
> 100 people involved), it is difficult to convince investors that betting on
> some obscure technology (be it a programming language, an ASIC,
> cabling, or what have you) is actually going to give measurable benefit
> in terms of faster time to market, lower cost, higher quality.
> After all, the name of the game for large companies is _not_
> programming brilliance, but being predictable and responsive in
> the eyes of large customers, and being able to deliver and support
> product over many years all over the world.
ie, the name of the game is risk removal.
> Again, many of these problems are specific to large companies, and
> Graham's "Beating the averages" holds that small companies cannot
> afford average performance, and thus have to be better and smarter
> than the big companies. By avoiding many of the pitfalls of choosing
> niche technologies, they may "safely" choose them and benefit from
> the higher productivity of using sharper tools.
> I saw a video of Steve Ballmer talking to people at (I think) Harvard.
> He was talking about the four phases of a company:
> 1) You have an idea, and try to bring it to market
> 2) You manage to get foothold, and milk it for all it's worth
> 3) You try to establish a culture for coming up with a new (1)
> 4) You recurse.
> He said most companies never get past (1), while large companies
> are typically doing (1)-(4) simultaneously.
> We often talk about programming language selection as if it were
> all about (1), but if one could do a real life-cycle analysis, it might
> even impress someone.
REPL for companies :-).
More information about the erlang-questions