Meyer, OO and concurrency

[Ulf Wiger]

Den 2005-07-14 23:00:14 skrev todd <todd@REDACTED>:

> Ulf Wiger wrote:
>
>> I don't think that's what they're asked to do.
>> Most of the time, what we ask people is to at
>> least _study_ the issue. Quite often, Erlang is dismissed
>> based on prejudice and, at most, a very cursory glance.
>
> At a telecom company I worked at we studied the issue.
> I was the erlang proponent.
> It may be that people do study, but don't come to the conclusion
> you think obvious because they value something different than you do.

It may be. Certainly some people will have studied the issue,
and not all will reach the same conclusion.

If your OS is VxWorks, for example, and you feel that using
semaphonres is a good way to solve your problem, then there
are loads of reasons why Erlang might not be a good choice.
We have actively discouraged people from using Erlang on
VxWorks in the past. I guess VxWorks supports memory protection
now, so combining C and Erlang on VxWorks is perhaps no longer
the disaster it used to be.

>> If you're investing millions annually into developing complex
>> software, why is it unreasonable to put some money into actually
>> _investigating_ if new implementation techniques might shave off
>> a portion of that, perhaps getting you to market faster, and
>> allowing you to save money on maintenance as well?
>
> Perhaps is the big word. Unless it's a clear win people want to stick
> with what they know will work.

Except some 7 out of 10 large software projects fail (or even
more, depending on which study you want to believe in most)

>
>>
>> Mike Williams used to say: "if you don't conduct experiments
>> at the beginning of your project, your whole project will
>> become an experiment".
>
> Every project is an experiment. Which is why people want to
> stick with what has worked in the past. Conducting a little
> experiment won't change the weight of experience. The interesting book,  
> paradox of
> choice,  talks about how people are risk adverse and would rather
> take a smaller win now than a potentially bigger but less sure win
> later.
>
>>> Don't trivialize what is involved.
>>
>>
>> At some point, many of us have had to choose: keep pushing
>> or take a step back and quit trying to make people see the
>> light. This is not trivialising the issue (well, to some extent
>> it is, but only relatively)
>
> When you say the reason people don't want to use erlang is simply
> because they don't want to use multiple languages,
> that is trivializing the issue.

But no one suggested that.
Joe suggested that one important reason why people dislike
mixing languages is that many language environments don't
mix well, and it's very difficult to get them to interoperate.

There are several domains where Erlang isn't a contender at all.

There are some domains where Erlang might be a good alternative,
but it's not obviously better than other, more established
approaches.

Then there are domains where Erlang is so much better than e.g.
C++ that practically all who give it a chance find it hard to
believe the grudgery they used to put themselves through.
Even in these domains, many people will refuse to give it
a chance. Some will claim that they do this on good grounds,
and talk portentously about business cases and risk reduction.
Some will fight it for political gain, or perhaps because they
happen to dislike a person who is much in favour of it.
Then again, some people will simply prioritize differently
and/or disagree that Erlang would be so much better.

(If I had to guess, I'd place you in the middle category.
Many of the arguments in this thread have been about the
latter one.)

Still, with very few exceptions, people are not malicious
and honestly try to make the best choices in each situation.
The fact that people given the same data can still draw
completely opposite conclusions is both stimulating and
good for our evolution, I think. (:


/Uffe