@council; cross posting to provide the reasoning, if discussion continues on 
council ml, kindly cc me (unsubscribed long ago).  Technical 
discussion (which should be the basis of "why it was banned" should be 
on dev ml imo).

On Tue, Apr 24, 2007 at 09:11:44PM +0200, Danny van Dyk wrote:
> Hi all,
> 
> [CC'ing council@g.o as requested by GLEP amendment from March 8th, 2007]
> 
> A subset of council members decided today that multiple version suffixes 
> are illegal in the tree pending further notice. This decission can be 
> appealed at the next Council meeting. If there is sufficient public 
> demand, an earlier meeting can be held.

Rules for 'appealing' are a wee bit sparse, but consider this email an 
appeal to reopen the issue at the next council meeting (and a 
suggestion to figure out what appealing requires/involves).  Offhand, 
while there has been sqawking, the functionality has been available 
for over a year (first 2.1 release of portage), pkgcore has long term 
supported it, paludis will support it in next released version (it's 
in trunk at least), PMS has the basic comparison rules doc'd out in 
addition.

As others have said, but reiterating in this message- the only 
'recent' change for multi-suffix is unlocking it in repoman so folks 
could use it; nature of backwards compatibility, the support had to be 
left locked for >6 months to preclude issues from stage releases, only 
change this side of 2007 was unlocking it.

Meanwhile, bug involved which is basically resolved at this point-
http://bugs.gentoo.org/166522

If the intention of the subset was to limit things till the allowed 
permutations of multi-suffix are worked out, please clarify- at least 
what I've seen thread wise, haven't seen a real explanation for it 
beyond "multi-suffix is icky and robbat2 has a hackish alternative" :)


> This decission has been made to prevent sufficient precedence for 
> unilateral changes to the tree structure. So far the following package 
> versions are considered illegal:

Please expand further on this one- no offense meant, but the 
offered reason is slightly weasely in that it's not really saying 
anything, what it is saying is pretty obfuscated.

Best I can figure, the offered reason is "it needs to be blocked 
before it becomes widespread thus cannot be blocked any further"- 
which isn't much of a reason since the support is long term there 
already, and doesn't state *why* it needs to be blocked (just states 
"it needs to be blocked").

I'm not a mind reader, so lets just assume I'm misreading it.  Either 
way, feel free to expound on the 'why' (either ml or via council 
appeal).


> An illegal version specification of media-sound/alsa-driver has already 
> been removed from the tree.
> 
> I would like to ask the affected package maintainers to move these 
> versions to sane version specifications as soon as possible. Thanks in 
> advance for this.

In the future when a subset (or full council, whatever) decides to ban 
functionality such as this, strongly suggest they ban *further* usage 
of it- implicit there is that the existing usage is left alone till a 
full decision can be reached.  Y'all banned all usage of it, meaning 
people have to make changes now.

Reasoning is pretty simple; at least for the two versions above, via 
making it illegal it forces them to transition to a hasty versioning 
scheme that may (frankly) suck- such as robbats proposal (his proposal 
works, but it's not human friendly and frankly serves more as a 
demonstration of why multi-suffix is useful).

Joking aside, if the intention is to block further usage till the 
permutations allowed are ironed out, fair enough- would strongly 
suggest not decreeing "they've got to go now" when you're stating in 
the same breath the decision will (effectively) be revisited a few 
weeks later.  Especially since changes to the versioning
scheme can be a royal pain in the ass transitioning away from 
afterwards.

~harring