[lsb-discuss] Looking for comments on Perl, Python (and why not TCL, Ruby, PHP, etc.)

[Andy Armstrong]

(background - I signed up for this list when Stew Benedict posted on
Perl 5 Porters raising the possibility of including Perl in LSB. I'm
from - but not speaking for - the Perl community)

On 5 Jun 2007, at 22:48, Wichmann, Mats D wrote:
[snip]
> The normal LSB model is to provide a functional specification of a
> feature - that is, a detailed specification of how the feature works,
> allowing any implementation which meets that functional specification
> to be considered conforming. Distributions then promise to continue to
> provide a functionally equivalent version of the feature, and tests
> are provided to make sure this is true. It's becoming increasingly
> clear that this level of detail will not happen for Perl and Python,
> definitely not in the short term and probably never.

That's certainly true for Perl. I think Python (multiple
implementations, simpler language) may be easier to nail down - but I'm
not a Python expert.

Perl is legendarily unspecified. Perl has always been defined as
whatever perl understands.

So I understand and agree with your concerns.

[snip]
>> From an ISV perspective, what problems are coming up with
> Perl and Python as things stand now, and in what way would those
> languages being in the LSB alleviate those issues?

If a machine has Perl installed that Perl will have already passed a
rigourous set of tests. Perl 5's test suite currently makes 64,793
assertions[1].

So pragmatically - as you suggest - it's unlikely that people are
experiencing problems that including Perl in LSB would fix.

[1] http://www.oreillynet.com/onlamp/blog/2007/05/trust_but_verify.html

-- 
Andy Armstrong, hexten.net