Daniel,

On Sun, Jul 25, 2010 at 7:37 PM, Daniel Carrera <[email protected]> wrote:
> This might be a stupid question, why is it so difficult to make a
> static distribution? I was hoping it wouldn't be too hard.

I'd agree that it should be easy, but in practice, there are two major
stumbling blocks:

1) With the combination of OSes and processors, we would end up
supporting approximately a dozen or so static builds, each probably 1
Gb in size, as we have to build a perl from scratch for each
combination. I know this because I did this for Mac OSX as an
experiment prior to building SciKarl in January 2010.

I also had to install Matlab on my machine. The Matlab install was
particularly illuminating - it took about 10 web pages of selections,
and one hour of installation downloads to get *basic* matlab
installed. Remember, this wasn't building anything, this was just
pulling down the binaries on a fast connection. Matlab and IDL have
full time paid people dedicated to testing and debugging these build
processes. We're volunteers donating our time to PDL - and thank
goodness for the great people on this list who do so willingly!

2) More fundamentally, PDL is built on Perl, and people expect PDL to
work with Perl modules they've installed themselves. WIth a static
build, there would be two perls on a given system, leading to lots of
Module confusion (I've tied myself in knots with this problem....)

We also have PDL dependent on libraries that are beyond our control,
such as fortran for PGPLOT, and I've always hit stumbling blocks with
PLplot not correctly building, and then people (incorrectly) blame PDL
for  these problems.

But, I could be wrong about this - if there's a magic way of building
just one static PDL for Windows and one for Mac OS X, I'm all ears!

Cheers,

Matt

_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl

Reply via email to