• janderson (3/3) Jun 03 2008 A friend at work pointed out this article to me:
• Sean Kelly (15/15) Jun 04 2008 This is a fairly interesting blog as well:
• renoX (17/22) Jun 08 2008 Very interesting, thanks.
• Robert Fraser (7/11) Jun 08 2008 When developing in Java for Eclipse (that has a ~1 min startup time), I
• bearophile (4/5) Jun 09 2008 It's not just a matter of syntax, Scala has other kind of ideas too. I t...
• Sean Kelly (21/43) Jun 09 2008 This isn't always feasible. Consider a phone switch (the purpose for
• renoX (10/16) Jun 09 2008 Sure, but straightforward isn't the same as easy to learn: you cannot
• Sean Kelly (20/34) Jun 09 2008 I agree that it's important for mutability to be available if
• Juan Jose Comellas (23/49) Jun 14 2008 Hot code swapping is essential for a lot of projects, especially if you ...
• Charles Hixson (6/30) Jun 14 2008 Isn't Erlang extremely slow, though? The comparisons I seen all place i...
• Robert Fraser (7/38) Jun 14 2008 For multithrreaded tasks it does much better, but not as well as
• Georg Wrede (4/7) Jun 10 2008 This is what I'd call invisible advertising in blogs. The mention of D
• janderson (5/13) Jun 12 2008 I didn't see a reference to D. Note: My motivation was that a wide
• Robert Fraser (2/3) Jun 13 2008 Quote at the end of the article.

janderson <askme me.com> writes:

A friend at work pointed out this article to me:

http://yarivsblog.com/articles/2008/05/18/erlang-vs-scala/

-Joel

Sean Kelly <sean invisibleduck.org> writes:

This is a fairly interesting blog as well:

http://www.sauria.com/blog/2008/05/28/notes-on-a-history-of-erlang/

It's largely a collection of quotes from Joe Armstrong, the creator
of Erlang.  Here's one of the many that may be interesting regarding
D 2.0:


family is dubious. Erlang programs are not referentially transparent
and there is no system for static type analysis of Erlang programs. Nor
is it relational language. Sequential Erlang has a pure functional subset,
but nobody can force the programmer to use this subset; indeed, there
are often good reasons for not using it."

The blog also notes that Joe Armstrong has never mentioned the Actor
model in relation to Erlang, and suggests that this means that the Actor
model was rediscovered independently during the creation of Erlang.


Sean

renoX <renosky free.fr> writes:

janderson a écrit :
 A friend at work pointed out this article to me:
 
 http://yarivsblog.com/articles/2008/05/18/erlang-vs-scala/
 
 -Joel

Very interesting, thanks.

Ssome advantage of Erlang are mostly theoretical in my book: 'hot code 
swapping' bah, if you want to have full availability you need redundant 
hardware and you can use this for software version change (though it's 
not trivial I agree).
If you don't need full availability you just stop your software for 
version change: it's much more simple this way.

Also, I disagree about the 'ease' of learning Erlang: Erlang has an 
alien syntax (which I don't like much) and an alien way of doing things 
(single assignment) which is not so easy for developers..
IMHO developers learn Erlang because it's good at distributed 
programming not because of the language itself whereas Scala has a 
really nice syntax (better than D's), but it's not as good as 
distributed programming some criticism in the article where interesting 
(cooperative scheduling)..

renoX

Robert Fraser <fraserofthenight gmail.com> writes:

renoX wrote:
 Ssome advantage of Erlang are mostly theoretical in my book: 'hot code 
 swapping' bah, if you want to have full availability you need redundant 
 hardware and you can use this for software version change (though it's 
 not trivial I agree).

When developing in Java for Eclipse (that has a ~1 min startup time), I 
frequently use hot-swapping and am often annoyed the sun JVM only allows 
hot-swapping for method body changes.

I looked into adding native code hot-swapping support on top of DDL, but 
it looked like all the challenges of writing a static linker + a dynamic 
linker + a compiler, so I gave up.

bearophile <bearophileHUGS lycos.com> writes:

renoX:
 whereas Scala has a really nice syntax (better than D's),

It's not just a matter of syntax, Scala has other kind of ideas too. I think D
may enjoy some ideas coming from Scala. If you (or someone else) knows Scala,
then I suggest to write a list of ideas that can be interesting for D.

Bye,
bearophile

Sean Kelly <sean invisibleduck.org> writes:

== Quote from renoX (renosky free.fr)'s article
 janderson a écrit :
 A friend at work pointed out this article to me:

 http://yarivsblog.com/articles/2008/05/18/erlang-vs-scala/

 -Joel

 Very interesting, thanks.
 Ssome advantage of Erlang are mostly theoretical in my book: 'hot code
 swapping' bah, if you want to have full availability you need redundant
 hardware and you can use this for software version change (though it's
 not trivial I agree).

This isn't always feasible.  Consider a phone switch (the purpose for
which Erlang was created in the first place).  In such instances, even if
the programmer has the money and the space for duplicate hardware
(unlikely, if you've ever seen a phone switch) then they would still have
to reroute all new calls to the backup box and then wait an indefinite
period for all existing calls to complete before cycling the old box.  As
someone who has actually done this, I can tell you that I'd have much
preferred simply having new calls simply be able to use the new software.

 If you don't need full availability you just stop your software for
 version change: it's much more simple this way.
 Also, I disagree about the 'ease' of learning Erlang: Erlang has an
 alien syntax (which I don't like much) and an alien way of doing things
 (single assignment) which is not so easy for developers..

I disagree.  The Erlang syntax is really pretty straightforward.  In fact,
the Erlang book covers the entire language in about the first two
chapters, and leaves the rest for discussing concepts like concurrency.
I felt that this was more than enough coverage for the language itself,
and was able to get going with Erlang in the short time it took me to
read these first few chapters.  There are some things that most users
won't need to know anyway, like bit packing and dealing with binary
blobs.

 IMHO developers learn Erlang because it's good at distributed
 programming not because of the language itself whereas Scala has a
 really nice syntax (better than D's), but it's not as good as
 distributed programming some criticism in the article where interesting
 (cooperative scheduling)..
 renoX

Fair enough.  I was drawn to D for its syntax rather than what it did as
well, so I can see how someone might be drawn to Scala for the same
reason.


Sean

renoX <renosky free.fr> writes:

Sean Kelly a écrit :
 == Quote from renoX (renosky free.fr)'s article

[cut]
 Also, I disagree about the 'ease' of learning Erlang: Erlang has an
 alien syntax (which I don't like much) and an alien way of doing things
 (single assignment) which is not so easy for developers..

 
 I disagree.  The Erlang syntax is really pretty straightforward.

Sure, but straightforward isn't the same as easy to learn: you cannot 
make a syntax more straightforward than Lisp's syntax but this doesn't 
make the language easy nor pleasant to learn..
Why? Because it's different from what we're used to.

And what about the single assignment?
Personally I'm not fond of functional programming only.

Regards,
renoX

Sean Kelly <sean invisibleduck.org> writes:

== Quote from renoX (renosky free.fr)'s article
 Sean Kelly a écrit :
 == Quote from renoX (renosky free.fr)'s article

 [cut]
 Also, I disagree about the 'ease' of learning Erlang: Erlang has an
 alien syntax (which I don't like much) and an alien way of doing things
 (single assignment) which is not so easy for developers..

 I disagree.  The Erlang syntax is really pretty straightforward.

 Sure, but straightforward isn't the same as easy to learn: you cannot
 make a syntax more straightforward than Lisp's syntax but this doesn't
 make the language easy nor pleasant to learn..

Fair enough.

 Why? Because it's different from what we're used to.
 And what about the single assignment?
 Personally I'm not fond of functional programming only.

I agree that it's important for mutability to be available if
necessary.  But Erlang integrates reasonably well with C,
so I don't consider this an issue really.  In fact, I prefer this
approach to the "all in one" approach that D 2.0 seems to
be aiming for, as I prefer that there be a semantic separation
between my functional and imperative code.  I feel that this
is a good way of preventing "bleed through" of concepts
that would complicate and confuse code.  It also neatly
avoids a motivator for producing monolithic code, and
the consequent risk that the mutable-state portion may
fail and corrupt the entire process.

But this is really a matter of preference.  I'm sure many
people disagree, and others simply don't work on the
type of applications Erlang was designed to target.  I
just happen to be in the niche that Erlang was specifically
intended for.  I only wish I'd heard about it 15 years ago
instead of spending all that time mucking about with C++.


Sean

Juan Jose Comellas <jcomellas gmail.com> writes:

Hot code swapping is essential for a lot of projects, especially if you need
to run 24x7, and it's completely trivial to do in Erlang. It's true that
its syntax looks alien and some of its limitations are baffling at first,
but there are very good reasons for them. Single assignment is very
important for handling concurrency. It makes the language safe and allows a
lot of optimizations at the VM level. In Erlang you normally have thousands
of lightweight processes (similar to Java's green threads) running, and
each one of them has its own heap, which can be garbage-collected
individually so that the application is never stopped completely during a
garbage-collection cycle.

Erlang is much simpler than either C++ or Java. The problem is just that it
doesn't share a "common ancestor" with them. That, and the lack of a huge
amount of libraries, are the biggest hurdles for its general adoption. I
still think that the future looks very promising for it, as there isn't any
other language that is better suited for tackling projects with massive
concurrency with the maturity that the Erlang VM provides today.

I've recently moved a 15-developer project from Java/C to Erlang/C and in my
experience it takes about a week for a Java developer to understand
Erlang's syntax and write simple programs with it, and 3 to 4 weeks to be
able to use Erlang and the OTP framework productively. It helps a lot if
you have been exposed to a functional programming language before, of
course, but it's not critical.


renoX wrote:

 janderson a écrit :
 A friend at work pointed out this article to me:
 
 http://yarivsblog.com/articles/2008/05/18/erlang-vs-scala/
 
 -Joel

 
 Very interesting, thanks.
 
 Ssome advantage of Erlang are mostly theoretical in my book: 'hot code
 swapping' bah, if you want to have full availability you need redundant
 hardware and you can use this for software version change (though it's
 not trivial I agree).
 If you don't need full availability you just stop your software for
 version change: it's much more simple this way.
 
 Also, I disagree about the 'ease' of learning Erlang: Erlang has an
 alien syntax (which I don't like much) and an alien way of doing things
 (single assignment) which is not so easy for developers..
 IMHO developers learn Erlang because it's good at distributed
 programming not because of the language itself whereas Scala has a
 really nice syntax (better than D's), but it's not as good as
 distributed programming some criticism in the article where interesting
 (cooperative scheduling)..
 
 renoX

Charles Hixson <charleshixsn earthlink.net> writes:

On Sat, 14 Jun 2008 12:26:43 -0300, Juan Jose Comellas wrote:

 ...
 renoX wrote:
 
 janderson a écrit :
 A friend at work pointed out this article to me:
 
 http://yarivsblog.com/articles/2008/05/18/erlang-vs-scala/
 
 -Joel

 
 Very interesting, thanks.
 
...
 
 Also, I disagree about the 'ease' of learning Erlang: Erlang has an
 alien syntax (which I don't like much) and an alien way of doing things
 (single assignment) which is not so easy for developers.. IMHO
 developers learn Erlang because it's good at distributed programming
 not because of the language itself whereas Scala has a really nice
 syntax (better than D's), but it's not as good as distributed
 programming some criticism in the article where interesting
 (cooperative scheduling)..
 
 renoX


Isn't Erlang extremely slow, though?  The comparisons I seen all place it 
several times slower than, say, Ruby at the same task.  (Granted the 
tasks were chosen to compare languages that are largely single stream of 
execution.  I'm thinking of places like 
http://shootout.alioth.debian.org/)

Robert Fraser <fraserofthenight gmail.com> writes:

Charles Hixson wrote:
 On Sat, 14 Jun 2008 12:26:43 -0300, Juan Jose Comellas wrote:
 
 ...
 renoX wrote:

 janderson a écrit :
 A friend at work pointed out this article to me:

 http://yarivsblog.com/articles/2008/05/18/erlang-vs-scala/

 -Joel

 Very interesting, thanks.

 ...

 Also, I disagree about the 'ease' of learning Erlang: Erlang has an
 alien syntax (which I don't like much) and an alien way of doing things
 (single assignment) which is not so easy for developers.. IMHO
 developers learn Erlang because it's good at distributed programming
 not because of the language itself whereas Scala has a really nice
 syntax (better than D's), but it's not as good as distributed
 programming some criticism in the article where interesting
 (cooperative scheduling)..

 renoX


 
 Isn't Erlang extremely slow, though?  The comparisons I seen all place it 
 several times slower than, say, Ruby at the same task.  (Granted the 
 tasks were chosen to compare languages that are largely single stream of 
 execution.  I'm thinking of places like 
 http://shootout.alioth.debian.org/)

For multithrreaded tasks it does much better, but not as well as 
multithreaded code written in a native language. When you're developing, 
say, for a supercomputer where a fifty clock cycle difference in an 
inner loop/frequently called function can man 3 more days of execution 
time when the program is used in production, these things tend to matter 
a lot more.

Georg Wrede <georg nospam.org> writes:

janderson wrote:
 A friend at work pointed out this article to me:
 
 http://yarivsblog.com/articles/2008/05/18/erlang-vs-scala/

This is what I'd call invisible advertising in blogs. The mention of D 
is quite brief, but amounts to product placing in film/TV, for which 
companies traditionally pay top dollar. Very effective.

janderson <askme me.com> writes:

Georg Wrede wrote:
 janderson wrote:
 A friend at work pointed out this article to me:

 http://yarivsblog.com/articles/2008/05/18/erlang-vs-scala/

 
 This is what I'd call invisible advertising in blogs. The mention of D 
 is quite brief, but amounts to product placing in film/TV, for which 
 companies traditionally pay top dollar. Very effective.

I didn't see a reference to D.  Note: My motivation was that a wide 
range of views/understandings are helpful in the never ending quest to 
improve programming skill and language design.

-Joel

Robert Fraser <fraserofthenight gmail.com> writes:

janderson wrote:
 I didn't see a reference to D.

Quote at the end of the article.