[TO]What's the big deal with EJB? [Re: PEP scepticism]

[Alex Martelli]

"Ville Vainio" <vvainio at karhu.tp.spt.fi> wrote in message
news:yoxd77fh4pr.fsf at karhu.tp.spt.fi...
> Mitchell Morris <mitchell.morris at cingular.com> writes:
>
> > All in all, it's been a pretty big win for Java. Conventional wisdom and
> > some non-peer-reviewed studies suggest that Java per se provides very
little
> > leverage over C++, that the average developer doesn't build Java
solutions
> > any faster or more bug-free than they build C++ ones, and that selecting
>
> Supposing that garbage collection and abundant reference semantics
> increase productivity by n percent (as all pythonians know), this
> would mean that C++ has some areas that improve productivity (and
> reduce bugs) as much over java. What are they? STL alone doesn't
> amount to that much, I would believe...

>From my personal experience: I get roughly a 15-20% increase of
productivity in Java.  Lack of templates (standard C++ library
and others) is BY FAR what I miss most in Java, reducing whatever
the gain would be if Java had them back down to such 15-20%
gains.  Other less-important lacks that hurt my productivity
in Java wrt C++ are multiple inheritance, auto-destructors
(having to code try/finally &c instead), and to some extent
slower recompile/rerunalltests cycles due to Java performance
in the second half of that cycle.

That's off the top of my head and based mostly on Java 1.1
experiences back when we were testing if it was the right
language for us -- what I was mostly doing was recoding in
Java the various proprietary network protocols, parsing and
loading data from old legacy file formats, etc, so as to allow
some tools (and eventually had we gone that way the whole
caboodle) to be coded in Java.  Small packages of a few
thousand lines' worth of C++ code each (say 2,000 to 15,000,
tops), roughly the same size in Java, so, no idea of how it
would fare in "programming in the large" tasks, GUIs, very
computationally-intensive tasks (handling network protocols
or parsing file formats is one thing, a 3D parametric geometry
kernel quite another:-), etc, etc.

Presumably the gain would be greater if the programmer did
NOT have full command of C++ with all of its complexities,
etc, etc.


Alex