|Think about Loose Coupling|
Re^3: Top Level Module Namespace Tableby Xiong (Hermit)
|on Feb 05, 2010 at 14:14 UTC||Need Help??|
Amazing how difficult it is for me to express myself precisely on this issue! I thought an example would be explicit. Sorry. Please allow me to attempt to improve.
I'm well acquainted with CPAN Search; it's a great tool. I'm not sure how much an ordinary dev will want to use Parse::CPAN::Modlist; but that's cool. Both of these are solutions to the problem of I want to download a module that does This. They are statements of what is. This is a common problem and an important one; but it's not the issue I'm attempting to address.
The table I seek is not descriptive but prescriptive. This is where the markers for deprecated namespaces come in.* This table goes no further than the top level of namespaces, perhaps some well-defined second levels (e.g., CGI::Application::). It's a statement of what should be.
Reviewing the descriptions of existing modules is indeed a way to infer prescriptions for future efforts, I admit. By the same token, one course of bricks is a guide to placing the next. Yet, if we want a wall to stand neatly, we generally use a plumb line.
I realize that I touch upon philosophical issues here.** My feeling is only that I will be more comfortable writing modules -- even modules that may never escape from my own project -- that are named in conformance with some plan.
I'm coming rapidly to the belief that no such table exists (I'd like to be wrong!) and that creating it is a worthwhile project. I wrote the original node in hopes that Monks might suggest a good way to get it started.
I don't personally assert that all modules currently under CLI:: should have fallen under Getopt:: or that all CLI-related modules should be created under the latter; but it's the kind of assertion that someone with more experience ought to be able to make.
I do assert that no module should be created under the Data:: namespace. I'm sure vigorous arguments can be mounted in support of it but I side with petdance that data is the world's worst identifier. All modules manipulate data.
Description vs Prescription is an old war in pedagogy, particularly among teachers of English grammar. At one time, prescriptive grammar was the unchallenged model, leading to textbooks insisting on awkward constructions and forbidding much common usage. The descriptive reaction reached its nadir with Ebonics.
I taught English as a Foreign Language for some time. My personal bias as an instructor is strongly descriptive; but in my own writing I try mostly to follow the ancient prescriptions. I do employ a few neologisms and nonstandard punctuation schemes. For instance, the rule is that quotes enclose trailing punctuation; I find this "intolerable".
My intention here is to balance, not overturn, the excellent descriptions provided by CPAN authors of existing work. I promise never to beat anyone over the head because his or her module naming doesn't conform to some standard; I don't have the stature. There is always the practical point that any module destined for CPAN may be submitted directly for review, at which time Wiser Heads can suggest the right name for it. (I do not suggest bypassing this step.)
The middle ground is often best. I would suggest that a writer of Perl, being an engineer, might be better served by prescription than the author of an English novel. Some will say both are artists and I don't disagree.