rfc: removing newnet from openrc
Ian Stakenvicius posted on Wed, 26 Oct 2011 14:55:45 -0400 as excerpted:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
> On 26/10/11 02:50 PM, William Hubbs wrote:
>> openrc has two network stacks currently. The first is the one most
>> people are using afaik, the net.* scripts, which I will call oldnet in
>> the rest of this message.
>> The second is the network and staticroute scripts, which we do not use
>> or support in gentoo, primarily because it does not allow the
>> flexability of the oldnet scripts. I will call these scripts newnet.
>> If there are no objections, I want to remove the newnet scripts before
>> the next release.
> It's been a while since I hung out in #gentoo, but one of the last times
> I was there (say, July?), there were people supporting the use of the
> newnet method (and i *think* were actually trying to get people to
> Personally, I prefer oldnet and would support the removal of newnet.
As with Ian I'd personally just as soon kill newnet, but...
AFAIK, this came up once before, during the stabilization discussion, and
each method had some users.
AFAIK newnet has never been officially supported in stable, and even in
~arch, it has been only the most forward leaning users that will have
switched, so IMO a news item isn't required as I'd otherwise suggest.
But a warning would still be useful, just in case, and I'd suggest a
deprecation/removal warning similar to that for tree-cleaning, 30 days
minimum, 60 days preferred.
I'm not sure what you mean by "release", 0.9.4, or 0.10.0, and the
planned release schedule.
If you intended killing it by 0.9.4, I'd suggest warning with that (and
do a stable 0.8.X-rY bump with the warning too, just in case, if 0.9
isn't in-process for stabilization already), with removal to be first
release next year (2012). If you do that within a week or so (both
upstream and gentoo), that will leave a full 60 days of warning.
If you were thinking about doing a 0.10 right away and keeping that
around for awhile, that could be more troublesome or less, depending on
perspective. I still think I'd try for a 0.9.4 (and 0.8.x-ry) right away
with the warning, delaying 0.10 a bit if necessary, and give it at least
that 30 days. With a full 0.x bump, users should be prepared for a few
more major changes, and a 30-day warning can be argued to be sufficient.
If you per chance were planning a 1.0, I'd say do it, without newnet, but
keep it masked for 30-60 days, during which the warnings can be running.
In that case, if appropriate, the warning can suggest unmasking 1.0 when
ready to upgrade, if they're still on oldnet or ready to revert back,
again, mentioning that it'll be unmasked on <date>, perhaps January 1,
That assumes that while newnet's not "supported", there's no existing
warnings about using it. If there are, and they've been there since
stabilization, then yeah, kill it, no further warning needed. (But, I'd
have thought you'd have mentioned it if that were the case, thus
assumption that it's not.)
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman