# from Ken Williams
# on Monday 12 November 2007 20:00:
>That doesn't jibe very well with the module case - otherwise we'd all
> be setting variables like $lib::Module::Build::VERSION .
Except bin isn't in @INC and foo isn't a .pm file.
But, require("./bin/foo") allows it to be loaded (assuming the
bin::foo::main() sort of structure.)
>In other words, 'bin' is not part of the name of the script, it's
>just where the script lives by default. I would rather do
>"program_name => 'foo'" and have it look for 'foo' in script/ .
I think we still need to tell it where to look, even if we decide that's
not part of the distro/package name.
Hmm, perldoc searches the $PATH, so technically there's already a
namespace collision between bin/foo and lib/foo.pm, though this gets
solved by 'man', right?
>> Would expect $bin::foo::VERSION and give you a
>> 'bin-foo-v0.0.1.tar.gz' tarball.
>
>Here too I think 'bin' shouldn't be present. What if you decide you
>want to offload some of the code from the 'foo' script into a module
>in lib/? And then expose some of the guts with APIs? The 'bin'
>moniker isn't really appropriate anymore.
True. I sort of see the the applicability of bin:: as being somewhat of
a frontend-only namespace though. As in: maybe it would be neat if
bin::foo->VERSION told me something about $0, but I guess that's what
$0 is for. Do we just go by $main::VERSION then? When your program
grows-up into a module, it can use the App:: namespace or whatever.
The guts could even be yanked into their own distro and made into a
dependency of the script-only distro (of course, just `mkdir lib` and
shipping with it would be fine too.)
I think we should be talking about keeping 'bin/' or 'sbin/'
or 'cgi-bin/' in the name in terms of possibly a META.yml field (which
would potentially solve the downstream packager issue that Adam
mentions.)
program_name: bin/foo
I guess it would become the debian package 'foo' (because they need to
address the namespace collision at the $PATH level.)
There is still a catch to addressability and namespace collision on CPAN
though. Because the index has nothing for scripts, I was thinking that
a "bin-foo" distro would hack that without interfering with any
pragmas. Unfortunately, I seem to be mistaken. See bin::sqlpp, where
a .pm file has been included apparently just to trick pause. So,
should we try to address this in pause, the indices, CPAN(PLUS).pm, or
just create a dummy .pm file?
--Eric
--
software: a hypothetical exercise which happens to compile.
---------------------------------------------------
http://scratchcomputing.com
---------------------------------------------------
