I think this idea is easy to do for a single target platform and harder to do for heterogeneous platforms.  For one thing you have current versions of Ubuntu shipping with perl 5.10 (and its variations in core). 

The available methods include:

Making a Bundle::Koha is probably the easiest and most achievable option.  The best evidence of this is that it has been done already:
http://search.cpan.org/~koha/Bundle-KohaSupport-0.08/KohaSupport.pm

Unfortunately that version is way out of date (since the author doesn't appear to consider versioning, perhaps Koha v1.0!) and it contains duplicates, but the idea is the same. 

The autobundle approach would be more of a shotgun method, getting stuff that isn't Koha-related and potentially limiting portability.

Building platform specific packages is the most difficult of the 3, but also the strongest solution for a given platform.  This would be consistent with your desire for "no CPAN", but not distro-agnostic.

I think the way to go is to make a current Bundle::Koha and let packagers build off of that.

--Joe Atzberger

On Wed, Jul 16, 2008 at 7:48 AM, Rick Welykochy <rick@praxis.com.au> wrote:
Chris Cormack wrote:

>> We cobbled together perl + all perl modules required.
>>
>> The virtual builder then simply installed perl and all the
>> perl modules we required.
>>
>> No CPAN. No failed makes. No mismatches.
>>
>> Can't the same be done for Koha?
>>
> Im sure it could, you volunteering Rick? :)

If in your opinion it would fly for Linux (as a start)
I wouldn't mind giving it a try.

I am familiar with care and feeding of Koha/2.2.9 but
have not ventured very far into Koha/3.

With 3.0 I got as far as prepping for system-wide Unicode
support, as that was highly recommended by the install
notes. I have not conquered Unicode on Linux/deb yet.
I would imagine it is important for universal acceptance.
(We have been spoilt by living in ASCII 7 bit land!)

That said, my guess for next steps is to examine the
Koha installation procedure for 3.0 and then unravel
all the perl modules required.

Then get some advice on rewriting the relevant parts of
the installation. I can do the perl stuff no probs.

Hopefully my solution would work with all linux distros
that have gcc installed.

Winders is another matter. I left Winders behind in 1998
so cannot directly comment, but my guess is that since
C compilation on Windows is a bit "tedious" for an out of
the box XP or Vista, that the Active-State precompiled
perl modules are the go.

Anywho, I could concentrate solely on the Linux way
of installation and then a Windows guru can adapt that.

Yeah, let's give it a go.

And git ... I'll need to (finally) learn some git. Not
a bad thing, imho.

Anyone else want to come on board as perhaps an advisor?

cheers
rickw

p.s. I just got a fresh new Asus EEE PC running Xandros,
so I might use that for starters, just because it sounds
challenging. Besides, my perl packaging idea should definitely
not rely on anything debian-specific (as but one example)
so the Asus is a good choice. I know nothing of its innards
yet and thus will not rely on anything distro specific.
I hope. In a word: it should be distro agnostic.

--
________________________________________________________________
Rick Welykochy || Praxis Services || Internet Driving Instructor