Mohan wasn't exactly talking to me, and I haven't exactly been
actively participating on this list in some time, but I thought I would
speak up with my opinion in this case... as GCJ has come to encompass
99% of my MinGW useage.
> So are you against the UNICOWS solution in that it seems to require
> the end user to adopt some sort of license for Win9X gcj-compiled
Yes, I would be 100%-passionately-near-violently opposed to it. I
still skim through the dev list on a regular basis, and some of the
things I've been hearing lately have alarmed me. If I've heard
correctly, I've heard at least one developer advocate for the GPL and
voice his welcome to the notion of introducing dependencies and
licensing restrictions. I've even seen a mention or two from someone
else (hopefully joking, I never know with Paul) about a POSIX layer down
the road. What the hell is going on?!?
MinGW has, and has always had, one and only one primary reason for
existence. That reason has been freedom from the licensing restritions
of Cygwin, period. Sure, "bloat" is a concern for many... but NOBODY
would have spent the past several years banging their heads against
their desks, trying to force code and makefiles designed for Cygwin to
work with MinGW, if they main concern was just to avoid having too much
of their hard drive taken up. Hell, most of my C++ education came from
trying to hack designed-for-Cygwin code to be compiled by MinGW... and I
did it all out of stubborn refusal to allow any compiler to dictate to
me the licensing terms of my end-products.
I like various open-source licenses, particularly the GPL. In fact,
every single line of code that I've ever publically release has been
licensed with the GPL. However, my original works have always been
licensed as such because of my choice, never by force. If the MinGW
team suddenly decides to deviate from this CORE FOUNDING PHILOSOPHY
after all this time, then what on earth will be the point of MinGW
anymore? I could just go back to Cygwin, which always has and always
will be a step ahead in development anyway.
The notion of introducing licensing restrictions into the MinGW
project is simply insane, and flies in the face of the past several
years of hard-earned reputation and goodwill. I beg of you to put this
discussion to bed and never repeat it again.
> If so, what alternative is the most palatable to you (for libjava)?:
> - pure ANSI without accented characters (which is what users are
> unhappy about now)
>- the UNICODE <-> ANSI <-> UNICODE which is inefficient for NT and
>- a configure-time switch allowing to choose among, say, ansi, unicows
> and unicode (i.e. NT targets with no Win9X support)
My first pick, at least until Win98/ME fade off to the sunset of
Win95 useage level, would be the second option. I know this may be a
contraversial thing to say, but lets be honest here... if the
what we were after, we wouldn't be developing on the Win32 platform to
begin with. Everything about this platform is a trade-off between
performance and backwards-compatibility, with backwards-compatibility
almost always winning.
I would be just as happy, though, with the third option. I really
don't see what the huge overhead would be, just a simple flag passed
into the configure script or makefile when building the compiler from
source. It doesn't sound like a maintanence nightmare, and it's not as
if SourceForge is charging us for the extra download hosting space.
Hell, call the NT-target-only build "Enterprise MinGW" and get cutesy
with it if you want.
Just whatever you do, keep licensing restrictions out of it. I
simply cannot believe that this current movement is entirely driven by
userbase demand, rather than developers' personal agendas. There is no
way that the general MinGW community would rate
optimal-Unicode-performance as being a high enough priority to abandon
MinGW's core principal of licensing freedom.