Re: Installing an Alternative Init?
On Sat, Nov 15, 2014 at 09:25:01PM -0500, Marty wrote:
> On 11/15/2014 07:45 PM, Ludovic Meyer wrote:
> >On Wed, Nov 12, 2014 at 12:26:26AM -0500, Marty wrote:
> >>On 11/11/2014 02:16 PM, Brian wrote:
> >>>On Tue 11 Nov 2014 at 12:36:14 -0500, Marty wrote:
> >>>
> >>>>On 11/11/2014 12:07 PM, Laurent Bigonville wrote:
> >>>>>
> >>>>>There are no functional differences between an installation with
> >>>>>sysvinit-core out of the box or an install where sysvinit-core is
> >>>>>installed later, this is a fact.
> >>>>>
> >>>>>Allowing the user to choose this at install time from the interface is
> >>>>>a "nice to have" feature (wishlist bug) not a RC bug like you were
> >>>>>claiming earlier.
> >>>>
> >>>>There is a potential practical consequence of not advertising an
> >>>>init alternative during setup. It makes users less likely to be
> >>>>aware of it, or even aware that the init system has changed.
> >>>
> >>>New users do not need to be be aware of all the background to the
> >>>choosing of a default init. No advertisement is needed. By definition,
> >>>they do not care. They want Debian. Please let them have it.
> >>
> >>They will not care "by definition" only if they are not aware of the
> >>change, and most won't be aware unless they are informed during the
> >>installation.
> >>
> >>>>They won't know they lost the choice they didn't know they had. Capisce?
> >>>
> >>>What choice have they lost?
> >>
> >>They lost an *informed* choice. I think the installation program
> >>should not take sides but just inform the user. A choice that the
> >>user is not aware of is the same as no choice, and is potentially
> >>coercive and disrespectful. It makes Debian seem partial to Red
> >>Hat's business plan to take over the Linux ecosystem.
> >
> >If you care so much about Redhat code, maybe you should document
> >yourself, and see there pay coders for glibc, gcc, the kernel ( a
> >ton of them, according to lwn and linux fundations reports ), on
> >coreutils, gnome, kde, php, python, openssh, etc, etc.
> >
> >>> Whatever it was, it didn't exist as you imply
> >>> in Wheezy.
> >>
> >>It wasn't an issue in Wheezy because the default init option had not
> >>changed from the previous release, and any release before that.
> >>
> >>>>They won't know, that is, until it bites them somewhere down the
> >>>>line. Then they won't know where to look or who to blame, and will
> >>>>blame Debian.
> >>>
> >>>What bites them?
> >>
> >>Individually, probably something that requires sysvinit or one many
> >>core services that got replaced. Collectively, getting trapped by
> >>vendor lock-in.
> >
> >You keep using those words, but you do not seems to use them correctly.
> >If the same system is present on more than one distributio, that's not
> >vendor lock-in since you can switch distribution and then reuse the same
> >system.
>
> I meant that one vendor seeks to control the Linux ecosystem.
> Whether that plan is viable or even sane, is another issue, but I am
> not eager to see if their plan will succeed or be a guinea pin in
> the experiment.
>
> (I would like to see systemd succeed in Debian, however, *without*
> sacrificing modularity or user choice. That would be like "embrace
> and extend" in reverse, and could serve to protect downstreams.)
>
> >Being tied to one package format ( and so on the ecosystem around ) would
> >be true lock-in. And no one complained that much since Debian existed,
> >despites the .deb having a few shortcomings at start, shortcomings that
> >were fixed later such as having checksum of installed software, a feature
> >rpm had at a time the dpkg didn't ( around 2002, so that's really a old stuff ).
> >
> >>In both cases it could be the result of users being steered to the
> >>default init by the installation program, leaving alternatives to
> >>rot.
> >
> >Alternatives will rot if no one use them, so either you recognize than
> >no one is interested to use them and it will indeed rot,
> >or that the few interested to use them are unable to fill bug reports and
> >help the alternatives survives.
> >
> >Given that a reading of the archives here show less than 50 people by a
> >large margin complaining on this list, I would indeed see that as a minority.
> >
> >( as I hope there is more than 100 000 to 1 million Debian users, since
> >Ubuntu speak of 20 millions, Fedora speaking around 2 or 3 millions. But that
> >doesn't matter, since 100 000 or 1 million, there would still be far less than 1%
> >of the user base ).
>
> I don't think Debian (or FOSS, for that matter) was ever about a
> winner-take-all approach to software choice. That seems to have
> originated in the commercial distributions, which have a financial
> interest in a) controlling users and b) controlling costs. I don't
> think that philosophy was ever part of Debian in the past. I had
> thought that all it takes is one interested maintainer to keep a
> package alive in Debian.
You are incorrect, on several point. First, it would be totally stupid to
want to control users when you give them the source code, way to build it
and the legal right to do what they want with it ( modulo GPL restrictions ).
You can still use the software after you stop paying for it in commercial distribution,
so if the goal is to control users, they are pretty bad at it. That would be like having a jail
without walls and no one to watch over.
In fact, I see more
that requests come from people excpecting others to do what they want, which
would be a form of control. People keep speaking of control as if it
was a point of discussion, but the root cause is that they do not themself
have any. And they do not have any because of the underlying social constructs
of free software, which is "coders/packagers have the power, the rest don't
until they become coders/packagers".
Now, the issue of controling costs exist for commercial and non commercials
groups, because the ressources are limited in both settings. Non
commercial entities also want to control costs, even if those costs are
not clearly expressed in dollars, unlike a regular company.
But if you have only 3 volunteers for your
house building NGO, you obviously have a limit on the number of house you
can build. The same go for any group that want to produce anything.
Thinking controling costs is only a consideration for commercial group
is being blind to the reality of any kind of groups.
> You might also be simplifying the problem. Software entanglement is
> a complex technical problem. There's a reason it's regarded as
> lock-in, because it's a technical cage that can be hard to break out
> of. It herds users in one direction, so the popularity issue is not
> only irrelevant, but misleading.
>
> I don't think there is a direct relationship between the number of
> users of alternate software, and the importance of maintaining it.
There is. The resources of maintainers are limited and I think we can totally agree
on this being a fact. Therefore, they need to prioritize the bugs.
While not everybody will doit in the same way, I think most developpers
will choose to take in priority bugs that affect more users, as seen by
the fact that hurd/kfreebsd and arm/mips/etc porters being the one
fixing the bugs on their side of Debian rather than the usual
maintainer. I think I can also safely say that there is way more
users of debian/amd64 than debian hurd on x86 or debian on mips.
Or you can also see the size of team maintaining some subsystem on Debian,
and how more popular software have more ressources to fix the bug than
less popular software.
Because statistically, the more users you have, all others things being equal,
the more will start to contribute and help, and so the more bug reports will be
corrected.
> I
> would say it's more of an opposite relationship, if user choice is
> valued. As less people use "locked-out" alternate software, those
> alternates arguably become more important to maintain to protect the
> choice of that minority. This of course presumes that user choice is
> still valued in Debian, which is something I no longer take for
> granted.
It was never granted, since you were always dependent on others
doing the work.
Now, if you want to make sure something is done, the best you can do is to
help doing it, ie, become dd, make bug reports, patchs, etc.
> >>>>Installation time may be only means that most users (like me*) ever
> >>>>would learn about it.
> >>>>
> >>>>* Install instructions? We don't need no stinkin' instructions
> >>>
> >>>Reading? You are right. Who wants it? Just spew out nonsense and hope
> >>>nobody notices.
> >>
> >>Isn't that where the dumbed-down install is headed? Don't worry
> >>about the details silly, Windows tells you when it's time to reboot.
> >
> >The part about Debian being a universal operating system also mean
> >it should aim for people who are not interested in details. Maybe you are
> >ok by having Debian being seen as "complicated and hard to use, spewing useless
> >questions on install", but that just mean than regular people will avoid it.
>
> It's not how I understood the concept of universal operating system,
> so that may be part of the difference of opinion. I've always seen
> Debian as a "proto-distribution" that downstreams can use to target
> specific demographics. I never interpreted it to target the common
> denominator desktop user. That is too limiting.
Sure, and how "proto distribution" is related to "universal distribution" ?
Distributions like openwrt or gentoo are made to be proto-distribution, and
are treated as such, and therefor have the technical infrastructure for that.
Debian do not really have the technical infrastructure for that ( for example,
do not have a central distributed VCS to let people easyly fork it as a whole,
do not provides a easy to replicate buildsystem so people can fork it, do not provides
useflags in dpkg like gentoo/openwrt, etc ), so if Debian really wanted
to be a proto distribution, it wouldn't do the best of job on it.
> >And if you want free software to be used, you would recognize that the setting
> >is advanced and do not belong to d-i.
> >
> >Now of course, maybe you are fine of having people staying on Windows or Mac OS X
> >because they have less trouble to install them and to use them, but you kinda
> >lose the right to complain " why do no one use Linux ?" ( and you also lose
> >the right to complain when others take that opportunity and are successful ).
> >
> >--
> >l.
>
> I see a place for Windows-like distributions. They are an old idea
> (remember Lindows?) It's just that I don't see Debian as filling
> that niche specifically.
"windows-like distribution" is a bit meaningless ( only because windows can mean anything
from windows server 2008 to win 95 or win 8 and windows CE ), and doesn't equal with
"distribution easy to use by non advanced users".
But since you disagree with the interpretation of "universal", I do not think
we can reach a conclusion any time soon.
--
l.
Reply to: