"Dave Thomas" <Dave / PragmaticProgrammer.com> wrote:

> We currently have a problem with Ruby under Windows. The Cygwin
> version shipped with the Installshield system is the most fully
> featured, but at the same time seems to be flaky in some situations
> (mostly due to Cygwin issues and the normal DLL hell). MinGW seems
> fairly solid, but there are rumors that it doesn't support all the
> background process stuff. The native VC++ version has less features,
> but is easier to integrate with third party DLLs such as Oracle.

One thing that has surprised me is that in my (admittedly limited)
experience, the cygnus ruby port has performed significantly faster
than either the mingw or the VC one. This observation is mostly based
on runs from one single script that does html-parsing through the raa
'sgml-parser' module. Throughout the 1.6.x series of interpreters so far,
the cygwin version has been consistently almost twice as fast as the
mingw one. I recently compiled 1.6.5 under VC, and the execution speed
ranked roughly between the other two ones. I find this surprising since
cygwin introduces an intermediary posix emulation layer which has to
cost something. I posted about this before but nobody seemed to have
an explanation. Did noone else notice such discrepancies? Of course if
this can help I'd gladly contribute the script to some benchmarking/
testbed for ruby under windows. Before rushing to nominate an "official"
win32 port, we should collect as many scripts/snippets as possible to
illustrate perceived stability/performance quirks in a repeatable manner.

Just my two (euro-)cents,

Bernard.