>>>> "Rick" == Rick Campbell
<rick(a)campbellcentral.org> writes:
Rick> Rick Campbell <rick(a)campbellcentral.org> writes:
> If you're distributing an executable over the net, it should
be
> statically linked, period. Disk and memory are cheap, throwing out
> program integrity to save disk and memory makes no sense.
Rick> However you forget that with modern libc it is close to impossible to
Rick> static link. c.f. Solaris.
Rick> Well, actually I didn't forget it exactly, but I did conveniently
Rick> forget to mention it :-) Still, wherever possible, one should
Rick> statically link for network distribution.
Rick> As far as I can tell, while Sun is fighting Microsoft in the public
Rick> arena, their technical decisions are some of the biggest reasons
Rick> around for switching to the dark side, er, NT. I'm not sure that even
There are people at Sun who have built careers out of transitioning
reluctant users from static to dynamic linking. At least for
libraries like libc, I agree with them. The real question is whether
you consider libc or system calls the real interface to the operating system.
I vote for libc.
Rick> HPs C++ compiler is as bad as Sun's is these days. It's really kind
Rick> of embarrassing to have to have a zillion backward compatibility hacks
Rick> so that code that works with all of the ANSI compatibility that comes
Rick> with Visual C++ can still deal with something as archaic as
Rick> Sparcworks(*). Meanwhile all their C++ effort seems to be in building
Rick> a friggin graphical development environment. Like, someone should
Rick> tell them that you're supposed to get the core functionality in place
Rick> before you starting playing with the bells and whistles. Jeez.
Rick> (*) Wherever possible, I try to use egcs, but too many management
Rick> types prefer to pay tens of thousands of dollars per license per year
Rick> for an inferior product because high quality open source products make
Rick> them nervous. Go figure.
With respect to C, the Sparcworks compiler is the most reliable
compiler I have ever used. GCC has been far worse in terms of
numbers of raw bugs encountered. I've rewritten major chunks of
XEmacs to avoid using buggy gcc features.
Dunno about C++.
Martin (who did not get paid for writing this)