[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Re: Legitimate exercise of our constitutional decision-making processes [Was, Re: Tentative summary of the amendments]

On Fri, 2014-10-31 at 09:37 +0530, Rustom Mody wrote:
> On Thu, 30 Oct 2014 16:54:33 +0900 Tristan van Berkom wrote:
> > [Disclaimer: I am not a debian developer myself and probably do not have
> >  the right to vote here, I am however a long time contributor and
> >  maintainer in GNOME who has been watching this thread and I feel I have
> >  a responsibility to add to this conversation ]

> 1. If the anti-systemd's brigade is really phobic and are not being
>    reasonable enough about "I dont want systemd anywhere near my system"
>    why not rename libraries like systemd-shim, libsystemd0 etc to not have the
>    phobia-inducing word?

I don't believe it's phobia, there are plenty of reasons, when
developing some embedded/single purpose product and using a reliable
upstream debian as your base, to choose the more tried and trusted
option, but let's not get into that (it's been done to death, and I
don't feel I'm contributing anything of additional value to this
conversation by going down that road).

> 2. Is systemd the problem or is gnome the problem?
>    On Mon, 20 Oct 2014 Charles Plessy wrote:
>    > Regarding what is proposed, it is actually unclear.  The consequence
>    > of accepting the main proposal may range anywhere between “do nothing
>    > special” and “harrass the GNOME and systemd maintainers until they
>    > quit”.  I am sure that this is not Ian's goal, but I am not sure he is
>    > in position to prevent this to happen.

I know I'm walking into a snake pit here but sure, I'll bite.

Is systemd the problem or is the GNOME Desktop Environment[0] ?

In my point of view the two now go hand in hand, they are both a valid
step into the future, however I don't feel that due diligence has been
taken for this technology's wide spread adoption.

The thing to understand here is that free software is generally slow in
terms of progress and excellent in terms of reliability, compatibility
and interoperability. This is the net result of having many entities
participate in software development in the open. Everyone has their own
valid requirements, so making sure software is developed properly in the
bazaar is much slower (and costlier) than it is in the cathedral, where
only one supporting entity's opinion matters.

So objectively speaking, what could have been done better ?

  o The GNOME Desktop needs D-Bus interfaces to bring event based
    notifications to the Desktop Environment.

    Instead of implementing a hurdle of platform specific code which
    manages to receive system specific events (with varying success),
    this D-Bus abstraction really is an improvement, this is the
    direction we should all be taking, we should be doing that together.

    But are these D-Bus APIs stable (yet) ?

  o systemd is currently the source of all of these D-Bus events which
    the GNOME Desktop requires.

    So systemd as an init system is possibly a nice young and upcoming
    init system, but now portends to be more than just that, with all of
    these extra services which replace components that we've already
    grown to trust.

    Did the GNOME/systemd maintainers propose patches to the already
    existing and trusted daemons to issue these D-Bus events ?

    Are there existing patches against udevd for instance, which allow
    udevd to provide the D-Bus events which GNOME requires for a good
    user experience ? (for a random example)

I think the answer to the above questions are currently 'no', which
means that the new GNOME Desktop was able to be developed much faster
than it would have been had they been playing fair and cooperating with
the wider FOSS community (which would require waiting on maintainers of
already trusted system components to accept, at least soft dependencies
on D-Bus and patches to produce the D-Bus events we require).

Again, I should reiterate that the approach RedHat has taken with GNOME
is a valid one, a lot of progress has been made quickly, and what has
evolved from that is a proof of concept/reference implementation which I
hope will be standardized appropriately and adopted by the wider FOSS

I don't think it's time to call this reference implementation stable, I
don't think that that would be a healthy decision for GNOME or for the
wider FOSS community.

As I mentioned in my initial mail, if one entity wants to foot the bill
for a development project as huge as this, we should not blame them for
not going the extra mile to meet the (heavy) expectations which are
appropriate in FOSS, however I also don't expect it to be considered
stable by other distributions so quickly, I would expect an incubation
period of 2-5 years or so while the D-Bus APIs in question stablize and
can be adopted by other Desktop implementations and other hardware

>    suggests that keeping the gnome-devs happy takes precedence over
>    other concerns
>    My impression is that when gnome went from 2 to 3 a large population
>    of debian-users (myself included) ran away... xfce,lxde, mate etc
>    Now the gnome-devs may be obstinate about their so-called progress but they
>    lost a significant user-base by their choices
>    windows lost with Vista, ubuntu with premature pulseaudio,
>    debian stands to lose by high-handedly pushing systemd down unwilling throats
>    [Note I am not talking facts but emotions and tempers:
>    It does not matter whether Marie Antonnete actually said "Let them eat cake"
>    The fact that that statement was attributed to her had bad consequences
>    for her and many others]

Much of what you write here ties into a controversy regarding GNOME 2/3
which has been done to death and I would rather not get into that.

> 3  Finally I see a lot of binary-thinking in the pro-systemd camp; viz
>    Either we push systemd as default or we carry around the obsolete sysv
>    forever.  I believe that all that is needed is an explicit policy
>     - for supporting sysv in jessie+1; not beyond
>     - if needed for now, make gnome an exception to that policy

iiuc, I think I agree with this basically, systemd/GNOME should probably
not be the default but rather the exception moving forward, at least
until this can be properly standardized and adopted properly.

Best Regards,

PS: I will try to refrain to write to this list any further, I do feel
    that I needed to say something as an upstream GNOME person who cares
    a great deal about the quality of the platform we produce in GNOME,
    but I really don't want to detract from the relevant discussion
    which Debian Developers are having here, and I'm sorry if my input
    has caused any unnecessary distraction.

[0]: I put a footnote here because I think calling the GNOME
     Desktop Environment 'GNOME' by itself propagates a sort
     of misconception. A lot of important and great software
     has evolved in the project which is GNOME, and only a portion
     of that is related to the 'Desktop Environment' use case.

     It's my opinion that the platform which has evolved in GNOME
     is of far greater significance than the flavor of the day
     desktop environment which is built with it, which happens
     to be the gnome-shell and it's components today, and maybe
     something else tomorrow.

Reply to: