[ruby-talk:4868] Re: Os there a cannonical Windows 1.6 distribution?

Andrew Hunt wrote:

#     >Well, first of all, *some*  users are going to be put off no matter
what
#     >you do, as long as trade-offs are involved. And "some" is likely to
be
#     >"many" in any case, although of course we want to make sure "many"
is not
#     >"most".
#
# We can make this simple then:
#
#     - A developer's distribution.  Small, just core Ruby.  You've got
#     to get Tk, cygwin, et al, all by yourself.
#
#     - A sumo distribution.  All singing, all dancing, everything you
#     could need including rdtool, rubyunit, etc.
#
#     - A user distribution.  No development tools per se, just the
#     tk/cygwin stuff needed to run ruby (DLLs only, etc.)
#
# What do ya'll think?

Well, I certainly think it is an important question. But first, the point
of the first part of my note was not that we should do different things,
but rather to concentrate on the best overall trade-off, with the
realization that many people will still most likely still be unhappy with
it, even at the minimum level of the unhappiness integral.

To follow up on the 2nd, unquoted part of my note, I think it is simpler
overall if you *don't* have multiple distributions. A major factor driving
the proposed SDK/sumo distributions is the realization that many things
(support, development, installation, use of applications) tend to be
easier overall if you have a fairly substantial and a fairly comprehensive
common frame of reference.

In evaluating size trade-offs, I think it is important to keep the rest of
the world in mind. Compared to the current Perl distribution and many Java
distributions, I think a standard full-fledged Ruby SDK distribution would
still be relatively small. For most people (and again there will still
likely be many, although hopefully a comparatively minimal number, of
unhappy exceptions), the marginal cost of storage space or download time
is almost completely insignificant, compared to "out of the box"
convenience (personal efficiency) factors.

Conrad Schneiker
(This note is unofficial and subject to improvement without notice.)