[STATSGRASS] R, gras, qgis etc in gentoo?

[Roger Bivand]

On Wed, 26 Sep 2007, Daniel McInerney wrote:

> Hi Agus,
>
>>  Sorry if this is too off-topic for this list.
>
> You might get more feedback about specifics
> of software on various linux distributions
> from the Freegis mailing list [1], as this
> list deals specifically with issues relating to
> GRASS-GIS interfaces to stats software, in particular R.
>
>>  I'm thinking on a new machine
>>  and wonder if gentoo linux would make
>>  difference for running R, grass, qgis...
>>  Anyone out there has any experience with or has
>>  information on this distribution
>>  in which all packages have to be compiled?
>
> To briefly answer your query, I would make the following observations:
>
> If your work entails monitoring system performance and how OS tuning will 
> influence processing speeds you might find that you have more control over a 
> system that is entirely installed from source, such as Gentoo. However, if 
> the objective is to do statistical analysis of spatial data using FOSS, a 
> distribution shouldn't make any difference - you should expect (& hope) to 
> reproduce the same results on all operating systems.

I think this sums things up well. As far as speed is concerned, you can 
make R go faster by using a tuned BLAS, because most of the heavier 
numerical code in R base will benefit from the tuning. For contributed 
packages, mileage will vary, if they use the same BLAS as R base, they 
will benefit too. So the differences are not at the distribution level 
really, given the same hardware, OS kernel, and compile train.

>
> My own personal experience is with Debian. All of the packages that
> you identify are available through APT and many more. You can also install 
> all of them from source if required. I'm not advocating one distribution over 
> the other, the choice is down to your needs, preference and experience.
>

I feel that compiling as many of the key software components as you feel 
necessary does help, because the build environment is (more or less) the 
same. The more components one depends on, the more likely it is that some 
should be compiled from source, say to keep all the bits using the same 
GDAL, rather than depending on different versions. The real benefit of 
access to source code is however being able to read it when necessary, to 
see what is going on.

Roger

> I hope that this is of use to you in deciding on your distribution.
>
> Best regards,
>
> Daniel.
>
>
> [1] http://news.gmane.org/group/gmane.comp.gis.freegis
>
>
> -----------
> Daniel McInerney
> School of Biology & Env. Science,
> University College Dublin,
> Belfield, Dublin 4.
> Ireland
>
> daniel.mcinerney @ ucd.ie
> ++ 353 1 716 7787 / 2698
>
> _______________________________________________
> statsgrass mailing list
> statsgrass at grass.itc.it
> http://grass.itc.it/mailman/listinfo/statsgrass
>

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no