Re: [Visualpython-users] Linux; VPython distribution

[Bruce S. <ba...@an...>]

--On Wednesday, March 13, 2002 8:28 PM -0500 art...@rs... wrote:
> But the performance on Linux (KDE
> desktop) seems quite weak versus Windows.  I suspect this has *nothing* to
> do with VPython itself - and am guessing it has something to do
> configuring Mesa and hardware acceleration.

Your diagnosis is probably correct. I have Red Hat Linux 7.2 (with Gnome 
desktop) dual booting on a laptop with Windows, and I find that VPython 
programs run twice as fast under Linux as they do under Windows!

> Firstly, it is normal in the distribution of Python modules to
> maintain the availability of  binary distros for versions of Python at
> least one or two versions back from  the current. Many people who might
> want to experiment with VPython but have no other compelling reason to
> move from, let's say, a happy Python2.1.1 install will simply choose to
> pass it by.

You are right that we should archive older versions for this and other good 
reasons; we will try to do that in the future. I have no interest in going 
backwards at this time to create packages for earlier versions of Python, 
but if someone wishes to create such packages I'd be happy to put them on 
the VPython web site.

> One of the beauties of VPython is its small footprint - a small *.dll and
> a few scripts and some docs are all that is necessary at the margin for
> anyone who already has/uses Numeric and who does not plan to use
> VPython_Idle.
> I understand that there are provisions for the separate download of
> VPython components - but I do not think it is clear to anyone coming in
> that there are good alternatives to the big footed standard distribution.

I don't see how this is a major problem. The "big footed" standard 
distribution may install Numeric again if you already have it, but so what? 
And it installs idle_VPython, but you don't have to use it. Since you 
mention "dll" you must be speaking of the Windows installer, which is not 
enormously larger as a result of including Numeric and idle_VPython (and 
remember that Numeric is needed by Visual). Where exactly is the harm? A 
slightly longer download time? However, see additional comments below.

> Though VPython no longer overwrites the standard Idle, it does, if I am
> not mistaken, overwrite the standard Python top level help index.

You are mistaken. The VPython installer (at least for Windows) copies the 
Python index.html to Python.html, then installs our index.html with a link 
to Python.html. So nothing is lost, and you have no evidence that we are 
sinning against Open Source manners. This kludge may well go away when 
current major work is completed by Stephen Gava to make IDLE configurable.

-----------------
Now for some background and philosophy. We found experimentally that even 
Carnegie Mellon computer science students had difficulties installing 
VPython with the early installation procedures, so we put major priority on 
all-in-one "big footprint" nearly foolproof installers, succeeding 
eventually on Windows in this respect. The Windows installer is basically 
nonproblematic now (except for archiving versions for earlier versions of 
Python).

We also went this route with the Mac, and there also this seems to have 
been absolutely essential because it is otherwise very difficult for a 
typical Mac user to assemble all the pieces. (And we're currently trying to 
tame the X11-on-Mac situation for VPython, which is a bear.)

It is really the Unix/Linux situation where the "big footprint" scheme 
isn't always appropriate, with the partial exception of Red Hat. What makes 
the Red Hat environment somewhat special is that relatively unskilled 
people come to Red Hat when they decide to try Linux, because to the 
general public it is the best known of the Linuxes. These relatively novice 
users are not well served by being given a shopping list of rpm's which 
they should find somewhere.

I should comment that work on better Unix/Linux installation procedures is 
relatively recent, and we are studying the advisability of using the Python 
distutils mechanism rather than our own script. But it is gratifying that 
you found that "I did a Python install on Redhat7.2 from the current distro 
and all went quite smoothly." Twas not always thus, and it is not an 
accident that it just happened to go smoothly for you.

That leaves a community not optimally well served: the knowledgeable 
Unix/Linux Python user who is perfectly capable of ferreting out all the 
necessary pieces and putting it together and simply incrementing an 
existing Python installation. We realize that we could do a better job of 
providing information to this group (without losing sight of serving the 
more novice community well, too).

> I understand that there are provisions for the separate download of
> VPython components - but I do not think it is clear to anyone coming in
> that there are good alternatives to the big footed standard distribution.

So it is mainly just a matter of improving/reorganizing the site 
documentation to make it clear that this is possible. The pieces are there.

Bruce Sherwood