Re: [Caml-list] Cross-compiling OCaml

[John Goerzen <jgoerzen@complete.org>]

On Tuesday 31 August 2004 04:05 am, Brandon J. Van Every wrote:
> james woodyatt wrote:
> > I'd rather that the Windows-centric guys on the list put their time
> > into good Windows support for Ocaml,
>
> No problem.  As I said, I realize Linux / mingw guys aren't into
> this.
>
> I'm just objecting to the statement that Linux cross-compilation
> support "would indeed be a great great additional functionality."  As

Well, let's look at that, because I believe you are missing the point.

Who ever said that cross-compilation support would only involve running 
a compiler on x86 Linux to target x86 Windows?  It could involve 
running a compiler on x86 Linux to target arm Zaurus, or m68k Linux, or 
x86 FreeBSD, or amd64 Linux.  Or running a compiler on PowerPC NetBSD 
to target Alpha NetBSD.  Or whatever else.

Believe it or not, this *is* useful.  In some cases, the target platform 
does not have enough resources to support a development environment 
(for instance, Arm-based PDAs).  In other cases, the target platform 
may not be available for the developers.  Or, it may be excruciatingly 
slow.  Or perhaps it is being bootstrapped and programs are being 
compiled for it for the first time.  Or perhaps it just sucks to work 
with (*cough* Windows *cough*).

> I think the reason you should care is because Windows is a big
> platform with a lot of users.  If you want to see the use of OCaml

Why should that make us care?  Why must you persist in measuring the 
success or failure of everything on pure user count?  I would say that 
is a pretty damn poor way to measure success, if not a completely 
stupid one.

> grow, so that there's more OCaml stuff available for all of us, and
> more paying OCaml jobs, then growth on the Windows platform is
> important. 

In that case, please explain the popularity of Perl, Python, sed, awk, 
Tcl, and Bourne shell.  All of which have had for a long time, or 
continue to have, roughly the same level of support for Windows as 
OCaml does.  Or less.

> Of course, some people don't have a platform-neutral world view. 

Actually, I think you'll find most people here *DO* have a 
platform-neutral "world view".

> Some people want Windows to die, more than anything else.  My own
> view is I just want platforms to be rendered irrelevant.  In the real
> world that means various engineering compromises, because platforms
> aren't the same.

Fine, but somebody HAS TO DO THE WORK to port things to such a 
different, expensive, and problematic platform.  It's a lot easier to 
port Linux code to FreeBSD than to Windows.  And a lot cheaper.

> Some of the archives I've crossed indicate that Cf may have no users
> at all, not just a lack of interest from Windows users.  Have you
> achieved a core of Linux users yet?  Nobody's going to bother to port
> stuff to Windows when the library hasn't proven its utility.

So you are saying that nobody on Windows is willing to try something 
new?  That they're only interested in "proven" technologies?  That it's 
only useful if it's popular?

In that case, you've convinced me to write off the Windows platform for 
MissingLib.  Thank you.

> Also, it helps to have a Sourceforge CVS project or the equivalent.
> http://www.wetware.com/jhw/src/ is digging.  You may actually be a
> very effective organizer, with wonderful source code.  But it doesn't
> look organized, it isn't publically indexed, it isn't publically
> source controlled, it isn't accessible in the way Sourceforge
> projects are. Also you have no webpage or mailing list for your
> project.

That's right.  It's accessible EASIER than SF projects are.  It took me 
about 2 seconds to get to what I'd want and download it.

With SF, it takes a lot longer.  First, I have to hope that SourceForge 
is up at the time.  Next, I have to find the appropriate project page, 
click on Files, wait for that to load.  Now, I get to click on a file 
to download and have yet ANOTHER page to wait to load.  There, I have 
to select a mirror, and finally I might possibly get a download if that 
mirror is reachable at the time.

I think that offering a simple tarball with the source is just fine.

> > and frankly— it's not like it really bugs
> > me.  It just tells me that Windows-centric guys don't like my code.
> > That's fine.  I don't like theirs all that much either.
>
> Your conclusion doesn't fit the available data.  The available data
> is your project has hardly gotten off the ground.  You have a
> recruitment problem.  You haven't solved it, because you haven't
> established basic infrastructure for such recruitment.

Although, according to you, the problem is that Windows people won't use 
something unless it's already popular.  So it's an insoluble problem 
for him, isn't it?

> Should I fault you for the public administration of your project? 

Oh please, tell us what is so wrong with posting the source code easily 
accessible as a tarball.  Is it that Windows people can't figure out 
how to use Winzip?

-- 
John Goerzen
Author, Foundations of Python Network Programming
http://www.complete.org/pynet

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners