Re: [Visualpython-users] History and Status?

[Bruce S. <bas...@un...>]

It is simple ignorance that I didn't know that one could produce a
self-executing file using distutils. I had only seen the type that asked you
to go into a DOS command window, get to the appropriate directory, and type
a command line. The availability of a doubleclickable installer would indeed
address one of the issues,  but there are still issues some of which happily
are being addressed by the (much) larger Python community.

The version of Idle created by Dave Scherer and bundled with VPython has
been the subject of intense further development, and with the blessing of
Guido is intended to replace the old pre-Scherer Idle in Python 2.3, which
would obviate the need for any special relationship between it and VPython.

Moreover, the new Idle has an excellent configuration scheme for referencing
help, which will make it unnecessary to tamper with the html documentation
files.

As to the larger issues, there is a tension between the norms of the
community of experts and the needs of nonexperts. Experts are used to
assembling applications from many sources. My views have been strongly
affected by observing experimentally what college students in engineering,
science, and computer science are and are not able to do reliably. What I
found was that it was very important to offer an installer (especially on
Windows, the platform most used by nonexperts) that bundled everything
together in an integrated way (Scherer's idle, Numeric, documentation).
Theoretically this could cause problems, although I've never actually heard
of any in practice, though one can construct a problem artificially if one
tries.

The only way I can see to address the needs of nonexperts, which
historically have hardly been addressed in the Python community, and also be
compatible with the expectations of experts, would be to offer two different
installation schemes: a bundle that does it all (this would for example
include Numeric) and a nonbundle where you install each piece (in
particular, Numeric would be listed separately). Since so far Arthur is the
only person who has complained about this, I haven't been able to make this
a priority, as it does involve extra work. Until very recently there has not
been sufficient people power to address issues such as this, but I certainly
don't rule out providing more installation options in the future.

Bruce Sherwood

----- Original Message -----
From: "Arthur" <ajs...@op...>
To: "Bruce Sherwood" <bas...@un...>; "John Keck"
<joh...@ho...>; <vis...@li...>
Sent: Sunday, February 16, 2003 6:54 PM
Subject: Re: [Visualpython-users] History and Status?


> > On the specific issue of distutils: It is probably correct that
distutils
> is
> > exactly the right mechanism for all Linux/Unix-like environments,
> including
> > the new Mac OSX 10.2 X11 environment. But it seems a poor mechanism for
> > nonexpert users on Windows, who typically don't have a compiler
available
> > (needed for source distribution) and are not comfortable working in a
> > command prompt window.
>
> You're under a misconception here.  Distutils makes a fine self executing
> installation file for Windows.  I have already prepared one for VPython as
a
> matter of fact - but it is VPython standalone, i.e including VPython demo
> and docs, but not including Numeric or V_Idle.  I personally see no need
to
> integrate the Numeric distribution with VPython.  Its own installation -
> including a Windows self installing executable - is quite expert, and I
> don't see it as a big deal to simply ask people who do not already have
> Numeric to download it seperately.  Unless I am missing something.
>
> In a way, this kind of thing is to me part of the issue. Its a form of not
> playing nice with other Python programs, IMO, to include Numeric in the
> VPython distro.  Numeric has thousands of users.  They should be able to
> download VPython without any fear of having their existing installation
> disturbed  in any way. For those who don't already have Numeric, its an
> extra click or two to install it as a separate distribution.
>
> If "we" want to attract other open source developers, I think we have to
> more carefully follow certain protocols.  I have felt strongly, for
example,
> that it is a bad  mistake to be manipulating the standard Python
index.html
> file. How might I convince you that is a signficant breach of protocol?
>
> Art