One of R's goals has always been to run on minimal hardware, and we say
for example (src/gnuwin32/CHANGES for rw1070)
This version of R needs more memory and is slower to start up, because
it loads more packages by default. This is only likely to be a
concern on machines with 16Mb of memory or less than 300MHz
processors. For such machines append R_DEFAULT_PACKAGES=ctest to the
command line in the shortcut used to run R, when the memory usage will
be about equal to that of rw1062.
Are people actually using R on machines with less than 32Mb of memory
or where it takes more than 10 seconds to start up, _and_ using a smaller
set of default packages to alleviate this? If so, please let me (not the
list) know.
If we can safely assume 32Mb and a 500MHz processor (where the startup
time for 1.8.1 is around 5 secs, less in R-devel) then we can make some
changes now, including allowing the use of S4 classes in R's basic
statistical functionality. (There are plans for 2004 that will load R
objects on first use and so both speed up startup and reduce typical
memory usage, but not until the second half of the year at the earliest.)
(Until recently I was using a 170MHz 64Mb Sun from 1997 but R no longer
builds on that machine.)
--
Brian D. Ripley, ripley at stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595
