Why Flatpak? Isn't Snap enough?

[Simon McVittie]

On Thu, 04 Jan 2018 at 16:26:06 -0800, Jasper St. Pierre wrote:
> On the Flatpak mailing
> list, people will argue that Flatpak's solution is technically better. On the
> Snap mailing list, they'll argue their solution is technically better.

Yes. It's almost as though mailing lists are somehow self-selecting? :-)

I don't have an answer for the fragmentation problem. It is a problem,
but I don't see how telling Flatpak people "you should give up on this
and contribute to Snap instead" is going to solve it, and I also don't
see how telling Snap people "you should give up on this and contribute
to Flatpak instead" is going to solve it. (Even, or perhaps especially,
if one of those is in fact the right thing to do.)

I don't think telling Flatpak or Snap developers "fragmentation is bad
and you should feel bad for being one of the sides" is necessarily going
to help either; people have chosen to contribute to that one and not the
other because they think it's a better answer, so obviously that one is
the one that should be the standard and the other should go away,
reinforcing the "us vs. them" problem.

I understand your frustration with Flatpak and Snap existing in parallel,
some apps having one, some apps having the other, and some app developers
feeling that they need to provide both. However, even if that situation
persists, it's still less fragmented than feeling that you need to
provide different RPMs for Red Hat, Mageia, SUSE etc., different .debs
for Debian and Ubuntu, the Arch Linux package format whose name I can
never remember, and so on; and at least it's possible to co-install
Flatpak and Snap on the same system and have them work, without being
forced to choose one or the other like you essentially have to do with
RPM/dpkg/etc. Two steps forward and one step back is still a step forward.

(I know I've written this as though Flatpak and Snap are the only
options for app distribution. That's only for simplicity of wording;
there are other projects in a similar space, like AppImage, and
essentially the same things apply with some of the names changed.)

> The social problem
> we have to solve is that people on the Flatpak mailing list would rather insult
> Snap rather than collaborate or pay any due respects

That's what I was trying to pre-empt by getting a more balanced answer
in first. Obviously, I failed. I'm biting my tongue trying not to say
why I think Flatpak is the better answer, because *obviously* I think
that, or I wouldn't be maintaining it in Debian; but it wouldn't
contribute anything meaningful to the discussion, so I'm trying hard
not to say it.

I think a certain level of fragmentation might be a necessary and
acceptable price to pay for deliberately trying to de-escalate the
"us vs. them" attitude. The fragmentation is unfortunate, but any time
a problem has several imperfect solutions with different tradeoffs,
there's going to be some competition between them. If we can encourage
it to be friendly competition with as much collaboration as possible
(like the way GNOME, KDE etc. share various freedesktop.org bits, and
the various Linux distributions contribute to the same upstream projects
they all use) then that's far better than shouting at each other.

Snap people seem to be interested in making use of xdg-desktop-portal and
my current work on app-container-specific servers in dbus-daemon (which
I started with Flatpak in mind, but aiming to make it general enough to
support Snap too), both of which are good bits of infrastructure-sharing.

Unfortunately, some sets of competing projects attract a vocal
echo-chamber of fans who don't contribute anything except advocacy,
which is occasionally helpful, but often degenerates into a negative
contribution by wasting the time of the people who do the work.

Not a reply to Jasper, just a general plea:

To anyone currently composing a message to a Flatpak or Snap forum
advocating Flatpak over Snap: please stop, delete the draft, and do
something positive to improve Flatpak instead. Fix a bug, or write
a detailed/useful bug report, or improve the documentation, or find
something in Flatpak's scope that Snap does better and work out how to
improve Flatpak until that's no longer true.

If people have already invested time and energy in Flatpak, your advocacy
is most likely not going to tell them anything they didn't already know;
if people have already invested time and energy in Snap, your advocacy
is probably not going to make them think "oh, I was wrong all along,
I should have been working on Flatpak" because that's not how the
sunk-cost fallacy works. If your advocacy insults Snap developers,
it's more likely to have the *opposite* effect by making them assume
the whole Flatpak community is equally rude. Please don't create that
impression.

Equally, to anyone currently composing a message to a Snap or Flatpak
forum advocating Snap over Flatpak: please stop, delete the draft, and
do something positive to improve Snap instead. The same reasons apply,
in reverse.

If one of Flatpak and Snap develops into an obviously better solution
for app development than the other, then no advocacy will be needed,
because it will already be obviously better. Writing flamebait emails
to mailing lists about how much better the other app framework is does
not improve the state of Linux software; making the other app framework
*actually better* does.

Thanks,
    smcv