Slashdot Mirror
Fedora QA Lead Pans Canonical 'Propaganda' On Snap Apps (happyassassin.net)
Long-time Slashdot reader JImbob0i0 shares a scathing article by Red Hat's Fedora QA "community monkey"/senior QA engineer on Canonical's announcement about their application delivery mechanism "snap"...
...and how it's going to unite all distributions and kill apt and rpm! This is, to put it diplomatically, a heaping pile of steaming bullshit... The press release and the stories together give you the strong impression that this thing called Snappy is going to be the cross-distribution future of application delivery, and it's all ready for use today and lots of major distributions are buying into it... The stories have headlines like "Adios apt and yum? Ubuntu's snap apps are coming to distros everywhere" and "Snap Packages Become Universal Binary Format for All GNU/Linux Distributions"...
Now, does Snappy actually have the cross-distribution buy-in that the press release claims (but never outright states) that it has? No... The sum total of communication between Canonical and Fedora before the release of this press release was that they mailed us asking about the process of packaging snappy for Fedora, and we told them about the main packaging process and COPR. They certainly did not in any way inform Fedora that they were going to send out a press release strongly implying that Fedora, along with every other distro in the world, was now a happy traveler on the Snappy bandwagon... They just decided to send out a wildly misleading press release and actively encourage the specialist press to report that Snappy was all set to take over the world and everyone was super happy with that.
Now, does Snappy actually have the cross-distribution buy-in that the press release claims (but never outright states) that it has? No... The sum total of communication between Canonical and Fedora before the release of this press release was that they mailed us asking about the process of packaging snappy for Fedora, and we told them about the main packaging process and COPR. They certainly did not in any way inform Fedora that they were going to send out a press release strongly implying that Fedora, along with every other distro in the world, was now a happy traveler on the Snappy bandwagon... They just decided to send out a wildly misleading press release and actively encourage the specialist press to report that Snappy was all set to take over the world and everyone was super happy with that.
Would greater consistency between the Linux distros make it an easier platform to target with binaries? Of course! But what we've seen happen with systemd, GNOME 3, and now these "snaps" is that this unification is happening around the worst technology around.
A typical Linux distribution today consists of the Linux kernel, which is actually quite good, and everything layered on top of that is shit.
We hear about systemd causing one problem or another far too often. For example, it was just a few weeks ago that we learned that a systemd change broke software like screen and tmux!
X is a trainwreck, but Wayland is even worse. With X we get a prehistoric relic, but at least it's somewhat usable. With Wayland all we get wheel spinning pointlessly in mud, going absolutely nowhere.
Then there's GNOME 3, which is by far the worst user experience we've ever seen from a Linux desktop environment. It's a tablet-oriented UI that nobody actually uses on a tablet, yet desktop users who try to use it are forced to endure the tradeoffs made to try to cram a desktop environment onto tablets.
People were attracted to Linux in the first place because it offered software that was better than what we found in Windows or Mac OS at the time. But that was 15 years ago, and times have changed. OS X and Windows provide a much better user experience these days for desktop users. FreeBSD is now indisputably better on servers.
These efforts to unify Linux have ruined it, I think.
"Unification" is exactly what GNU/Linux *doesn't* need. Innovation creeps up in the FOSS world through other projects, NOT through iteratively improving one core. I'm not a fan, but it's undeniable that systemd didn't become popular simply by building on initscripts. Even OpenRC doesn't do that; it uses sysvinit, but the scripts are entirely its own. Iterative development is not enough to create large breakthroughs. In the systemd example I gave it's really not that large because it doesn't actually do anything new, but it's still such a large change that it wouldn't be possible if it wasn't its own project.
In the "wonderful" unified Linux world, you'd be stuck with what you're given. If upstream doesn't want your contributions, you no longer have the (realistic) option to fork. You either get it into the mainline release, or your innovation or improvement simply doesn't get used because everyone's already "agreed" on this base.
Unification destroys the concept of distributions, destroys the ability to fork and build, destroys the ability to greatly change and innovate. I may not be a fan of some of the recent changes (such as the systemd cult), but I absolutely respect the shit out of the environment that makes that change possible or even easy. Whether you're running OpenRC, systemd, sysvinit, runit, or whatever, you're still GNU/Linux at the end of the day. The ability to pick and choose your software in Linux is a feature, not a bug. If you don't want configurability or choice, go use a BSD or literally any other OS.
The main issue with "targeting" is itself. Software should depend on as general a base as possible. For example, it shouldn't be using low level dependencies with unstable APIs. Most applications (games being the biggest examples) can get by with dbus, udev, xorg/wayland, and SDL. Other libraries such as libav, libpng, and so on are rather stable if you target them correctly.
If we want to point the fingers at anyone, it's the closed sources guys modifying the libraries that they use and expecting every distro to use their modifications. In the same breath, Ubuntu is one of the worst distros to target *because* they modify the libraries so much. The rest of Linux land doesn't suffer all that much, if at all. Build your shit right and it'll work.
Trying to force GNU/Linux land to unify will destroy its software ecosystem. Hackers will move on and/or simply refuse to target the unified stack. I've heard of some developers actively putting code into their applications to exit early if it detects software that the author disagrees with. Things could get ugly fast.
Did you actually try reading my post? Like, the bits with direct quotes from the Canonical press release?
Here, here's an easy one. The very first sentence of the press release claims Snappy is "enabling a single binary package to work perfectly and securely on any Linux desktop, server, cloud or device". *Present* tense. The claim is that it does this right now.
This is patently and demonstrably false. The snappy packages for Fedora are built with confinement disabled, and the installation instructions tell you to disable SELinux to use snappy. Thus the whole supposed 'security' feature of snappy itself is inactive (the snaps are not confined, they have full system access), and using snappy requires to to *significantly decrease* the security of the whole system (not just the snaps you're running).
Not to mention that meaningful confinement of X11 apps is impossible, and all major distributions still use X11 by default. This is why no-one is running around telling everyone they should use Flatpak right now. I presume it's also why Canonical engineers haven't been running around telling everyone to use Snappy right now. Canonical's PR department appears to have no such qualms, however.
So. There's a nice easy one for you. But for the advanced class - Canonical PR are not a bunch of idiots. They know exactly what they're *doing* when they issue a press release with a lot of key words and phrases carefully arranged in extremely ambiguous ways. You can bet your bottom dollar they were not shocked and amazed when all these stories came out saying that Snappy was now the agreed-upon universal app distribution system and apt and rpm would be dead any minute now. (Or if they were, it was in the "I'm shocked, shocked to find that gambling is going on in here!" sense of the word). They knew exactly what they were doing. Have you seen 'em running around issuing clarifications, fr'instance? No? Hmm, wonder why that is. It's fairly apparent from the articles that *were* published that Canonical held a press call to go with the press release; do you suppose they were carefully correcting the assumptions of journalists on said press call? Hmm? Doesn't look like it, does it?
Well. You may not see it this way, but to me there's a rather big difference. Usually when people talk about RH 'causing grief' and 'UNIX design philosophy' (sigh, if I had a nickel for every time...) they're talking systemd. Yes? Well, sure. Lennart wrote systemd, RH is fairly solidly behind it these days (though note it wasn't at first - Lennart had to sell systemd inside RH about as hard as he had to sell it anywhere else), and quite a lot of people don't like it.
Fine, it's a free world. But that's ultimately a technical argument. We didn't put systemd under an RH CLA. We didn't issue press releases prematurely declaring that it was taking over the free world. It's a freedesktop.org project, you don't have to sign your soul over to RH to use or work on it, it has plenty of non-RH contributors, and it got integrated into non-RH distros through their usual processes for feature review.
I don't usually actually have any problems with Canonical's engineers, or their projects, believe it or not. Of course there's the whole Wayland/Mir mess, but that's kind of an exception (and even there the main problem is down to management, not engineering). The stuff I don't like from Canonical invariably comes from management and/or PR, and (again purely my personal opinion) ultimately derives from Mark and his poor-man's-Steve-Jobs complex.
I don't have any particular problems with Snappy as a technology. Heck, a couple of things about it might be better than Flatpak (I don't know either system in much depth, just broad overviews and the specifics I dug into for this kerfuffle). From a purely technical viewpoint - if you ignore the publicity, and the problematic influence of snappy being under the Canonical CLA and the server end being a black box - it's perfectly possible Snappy could turn out to be the best answer to this particular question. It's certainly not a Wayland/Mir situation - Snappy and Flatpak both have fairly complex histories and predecessors, but whichever way you cut it, they've been around about as long as each other.
The issue I have is specifically with *this press release about snappy*, and more specifically with the way it vastly overstates snappy's current capabilities, and the way it strongly implies that snappy already has substantial cross-distro buy-in. Taken together - and if you look at the stories that came out of it, which Canonical PR *certainly* was not ignorant about, especially given there was a press call - this comes off as an attempt to effectively pre-empt the whole process of building consensus around a solution by giving Snappy such a strong press push that everyone just has to fall in line behind it, regardless of the fact it's not remotely *done* yet and there are other options that have already been trying to build support the right way.