Andreas J Koenig wrote:
I think I repeat myself: I am not intending to alter what the META.yml
files report to me. I will apply heuristics to find out if I can trust
the META.yml file and I will report back to the user if the META.yml
seems bogus to me. But if a META.yml is there and is reasonable, I
will gladly put its content into the index.

But before I can even start, I have to investigate how good or bad the
existing META.yml files are. So please be a little patient.

I'm not trying to rush you. I'm just suggesting that there *isn't* anything currently in the META.yml file you can use at this point, except for the package version. So there are two pieces:


1) You need to determine whether the existing META.yml structure meets your current needs;

2) We need to propose and develop an extended format, and both document them as well as preferably adding that generation to Module::Build and probably as a standalone (so ExtUtil::MakeMaker users won't be left out in the cold).

In the case of SVK, the current META.yml should, in fact, have the correct 
behavior:

name: SVK
version: 0.991
abstract: A decentralized version control system
author: Chia-liang Kao <[EMAIL PROTECTED]>
license: perl
distribution_type: module
...

so you could trust it for the base module's version, and the indexer would work again OOTB. I'd suggest that an appropriate extended structure would be:

version:
  SVK: 0.991
  SVK::Command: 0.991
  SVK::Command::Add: 0.991
  SVK::Command::Admin: 0.991
  SVK::Command::Annotate: 0.991
...and so on...

and the indexer could branch on whether "version:" was a scalar or a hash. In either case, the indexer should probably take the META.yml information as advisory only, and only fall back to it if the normal resolution method fails.

John

--
John Peacock
Director of Information Research and Technology
Rowman & Littlefield Publishing Group
4720 Boston Way
Lanham, MD 20706
301-459-3366 x.5010
fax 301-429-5747

Reply via email to