On Tue, Dec 2, 2008 at 9:01 PM, Aaron Griffin <aaronmgriffin@gmail.com>wrote:
On Tue, Dec 2, 2008 at 5:39 PM, Allan McRae <allan@archlinux.org> wrote:
This starts the official discussion period for the addition of rules governing the addition of packages to [community]. As this is essentially a bylaw change, we will use that voting procedure: 5 days discussion, 7 days voting, quorum of 75% required.
[proposal]
* Only "popular" packages may enter the repo, as defined by 1% usage from pkgstats or 10 votes on the AUR.
* Automatic exceptions to this rule are: - i18n packages - accessibility packages - drivers - dependencies of packages who satisfy the definition of popular, including makedeps and optdeps - packages that are part of a collection and are intended to be distributed together, provided a part of this collection satisfies the definition of popular
* Any additions not covered by the above criteria must first be proposed on the aur-general mailing list, explaining the reason for the exemption (e.g. renamed package, new package). The agreement of three other TUs is required for the package to be accepted into [community]. Proposed additions from TUs with large numbers of "non-popular" packages are more likely to be rejected.
* TUs are strongly encouraged to move packages they currently maintain from [community] if they have low usage. No enforcement will be made, although resigning TUs packages may be filtered before adoption can occur.
[end proposal]
I throw something in here during the official discussion period, directed at all the people saying "omg these metrics suck" (regarding pkgstats and votes).
The fact is, it's come to all our attention that we need _some_ way to control packages in community. These are the only metrics we have at the moment. The above bylaw, no matter what the actual metric used, is a decent one. Modifying the metric at a later time can and should be done, but for now these are the only metrics we have.
Simply put: some structure here is needed. If this is all we have right now, we should do it, rather than say "screw it, let's stick with this freeform thing we've been doing". If we find these metrics to be wanting, we can change it later. And seriously, how hard is it to go onto IRC or the forums and say "I want to put Foo in community, and need 8 more votes please!"
Well Aaron, even proponents of this proposal have said that the statistics and/or votes are a problem as far as accuratcy. Heck a simple examination shows that the voting system has over two orders of magnitude for a variable and NO basis for determining the shape of the possible distribution of votes , so no understanding of what then constitutes a mean either. AND you have again said it is needed to be doing this. So for the past several weeks many have been asking why. Would you PLEASE explain why this is needed. In detail, and with some statistics that can be verified and have a good basis in compilation. As it would be NICE to know exactly WHY this is NEEDED. And how this change accomplishes such a needed change. Best regards; Bob Finch