| [reply] |
Managing multiple different sourced programs quickly becomes chaotically complex. It's much better to maintain a single update source for major system components.
Or: Keeping track of when you need to update you own installed version of perl, as well as the system version of perl, is just more work than most people want to deal with.
Also: In case you are the sysadmin, it's fairly easy to accidentally replace the system perl with your new version. Which could have immediate or delayed problems. Immediate if something else in the system is broken by the new version of perl, delayed when your next system update removes the new version and puts the old version back (and then breaks some of your scripts...).
The last one, notably, doesn't apply to CPAN modules generally (there are a very few cases where it does...): If you install something that's not in the system, it won't affect anything in the system, as they will never use it.
And, depending on the situation, the developer may not have control over - or even access to - the deployment environment. Think of the case where they are developing something for a department, which will then be placed on multiple department servers.
Generally, I think of it as best to require as few auxillary software changes as possible when you are developing software. If there is something that really will make a difference in how long it will to get the program written (or if it gets written), use it, but if it would just be 'nice to have', try doing without.
Often that will make a difference in whether you can get approval to use the script you just wrote, or not.
| [reply] |
I don't get this argument. We always tell people to download the whole world from CPAN, even offering how to use modules if you can't install them the place your sysadmin does.
But going to CPAN and getting perl itself is somehow a big NoNo?
It's not so much a "NoNo" as potentially a lot of work for the testing staff and/or the sysadmins.
If you install a new module from CPAN, that module can't break your existing code, because there isn't any. No additional testing is needed to prove that you can safely promote your code.
If you need to upgrade an existing CPAN module, you will need to do some extra testing to in order to prove that the new version doesn't break any existing production code which uses that CPAN module.
If you need to upgrade perl itself, you need to do a lot of extra testing to prove that the new version of perl doesn't break any existing production code that uses perl.
Depending upon how much perl code is in production, this can be a lot of extra work. I've seen a new installation take several weeks to approve and complete.
Whether or not it's worth it to do that work is a business decision: based upon some kind of a cost/benefit analysis of the relative merits, costs, and risks.
| [reply] |
Your argument might have some sense if there was only one possible perl on a system.
But that isn't true. If it is an issue to upgrade /some/path/to/perl, then you can always use /some/other/path/to/perl. Or you call them /usr/bin/perl5.10.1, /usr/bin/perl5.11.3, /usr/bin/perl5.8.8 and have /usr/bin/perl be a link to whatever default you fancy.
Depending upon how much perl code is in production, this can be a lot of extra work. I've seen a new installation take several weeks to approve and complete.
But that's not a reason to assume that's true in general -- and I'm certainly not going to assume some kind of installment/upgrade policy when answering questions. Furthermore, organisations that have procedures in place for installing/upgrading new software usually have that for, uhm, new software. Whether that's perl, a C compiler, a CPAN module, or whatever their development department throws over the fence.
| [reply] |