nALFS programming details

Kevin P. Fleming kpfleming at linuxfromscratch.org
Mon Aug 9 16:23:41 PDT 2004


Joachim Beckers wrote:

> I was wondering whether we should think about making nALFS2 
> object-oriented. That of course requires that it would be written in 
> c++, and I don't think any of us is a real c++ programmer. (I try to be 
> one, but so far I've only read a couple of books about it :-)

Joachim gets the points for reading my mind this week :-)

I've been thinking along these lines as well, especially if SOAP is the 
chosen protocol (which I fully endorse) as it is very object-oriented 
itself.

I would differ with one statement here, though: choosing to implement 
nALFS2 using an object-oriented model does _not_ mean it must be 
implemented in C++. In fact, I've been considering suggesting to 
everyone that they start out building prototypes in some other language 
that's a little easier to work in, whether that be Python, Perl, Ruby or 
something else. That's what I've done working on major apps in the past: 
throw together a really ugly (but quick) prototype in a scripting 
language to work out what the inputs, outputs, protocols, etc. are going 
to be, then go back and start implementing for real in a more efficient 
(at run-time) language.

If a significant portion of nALFS2 gets implemented in, for example, 
Python, we may very well learn that we need such a thin object layer 
that even C++ would be overkill, and thus we could just implement what 
we need in straight C with well-defined coding guidelines instead.

Personally, I know no C++ at all, and am not keen on learning it. I do 
know enough of these other languages to be able to read them and help 
out, and could probably eventually do some real coding, but my skills 
really lend themselves to taking what someone else has already built and 
reimplementing it for maximum efficiency/safety/etc., which usually 
involves moving to a lower-level language as well.



More information about the alfs-discuss mailing list