[SciPy-dev] old web site front page

Joe Harrington jh at oobleck.astro.cornell.edu
Thu Dec 22 09:06:48 CST 2005

You and Ed came up with similar suggestions (on your first two
points).  Sure, take out the "VERY" and change "like" to "don't mind".
But, the more you guys argue that this is not a hard install, the more
you convince me that it is!  Arnd, you applied FIVE conditions just to
make it "pretty clean"!  But, the "pretty clean" builds have binary
installs already, leaving the "not-so-clean" and "decidedly rotten"
ones for trial by newbies.  Recall that this message is going to
people who have never installed SciPy and never used it!  We're not
talking to list members here, we're talking to the guy that walked
into my office the other day and said, hey, I hear you know something
about this free IDL for Python (I cringed).  Sure, he's typed "make
install" a bunch in the past.  But, I wouldn't wish a newcore build on
him, let alone a full build.  He's got work to do, and he's more
likely to walk away from this if it wastes his time than to dig in.

Is SVN better than tarball?  I'll leave that to a developer to say.
We can change it as the story changes.



The text reads very well!
FWIW, just some small comments:

On Wed, 21 Dec 2005, Joe Harrington wrote:



The *VERY* seems too strong to me -
Personally I would not call myself as "very experienced"
and still I celebrated the 100th build of scipy new core
(on one machine, where I could automate the stuff - no counters
on the others ;-)
Of course I have to admit, that I really learned a fair
bit during the last years of going through the various stages of scipy ;-)

Therefore I propose to omit the "VERY" to get more testing.

> like getting your hands dirty,

also this is maybe a bit too discouraging?
(I think this phrase could be just left out)

((if one uses gcc (<4.0), g77, a plain ATLAS and a normal fftw2
on a normal machine things are pretty clean already))

> download the tarball for the new version,

should really the tarball be recommended at this stage, or
wouldn't svn (or just http://new.scipy.org/Wiki/Download) as only pointer)
be sufficient?

More information about the Scipy-dev mailing list