Their server implementation has shades of their desktop implementation with snapd. It's also a sluggish performer. I haven't been happy with it starting with 16.04. Performance took a big hit and it has continued to degrade as I've been upgrading, though the 18.04 desktop is significantly better than the 16.04 desktop. I am actively looking for another distribution. probably non systemd based. I might even look at bsd. The new live installer is a snap based installer. It's broken, too.
It's bad enough that I'm even thinking about installing gentoo on something.
Snap is baked into the server. That's bad for all the security reasons listed by @lah7 above. The packages have bloat because the have their own libraries which will be out of date by the time you download the snap. It makes the snap a security risk right up front. I have a problem with any of these container based distribution schemes. They add too many layers. they make the entire network stack too complex in order to make them work at all.
As long as the user doesn't have the option to disable auto updates in favor of manual updates I refuse to use snaps on my system. The community have given feedback about this but the Ubuntu developers refuse to add that functionality (as is their prerogative). Instead of snaps I 've been using flatpaks when it makes sense (ie Viber, Skype etc). I 've personally moved from Ubuntu MATE to Fedora MATE because snaps are becoming more common place in the Ubuntu ecosystem and I don't think it's a good idea to fight the platform I use instead of finding a more aligned, with my use case, platform.
P.S. For anyone considering similar move Fedora MATE is not as well polished as Ubuntu MATE. For example not having MATE tweak makes a difference. The Ubuntu MATE team deserves congratulation for how well they 've implemented MATE DE on top of Ubuntu. Personally I don't mind the extra paper cuts when using my system for the trade off of being in control of my system (ie it updates when I approve of it). This trade off may or may not apply to other users.
The road to hell is paved with good intentions
I can understand the Intent of the Snap Package concept, as I understand, to make available or pool all apps produced for all distros for all Linux users. If that's not it then I'd love to understand what are all the intentions.
From what I read the Snap Packaging process is more of a development burden for developers. However, one must ask, if the effort of porting an application over from the deb family to the rpm family, or the reverse, and then maintaining both is not more of an effort in the medium and long run.
Is it also possible that this will add further stability to distros? If an app is packaged as a snap and tested on one distribution, the intent, my understanding, is it will work for all others. But is this certain? For how many upgrades?
What about the Snap packaging itself: if there are SSL, VM, drivers, GPL, or other improvements that come down the tube, how heavy will the packages become to stay upstream compatible?
What about security :
first there were VM to isolate environments from software bugs and malware, in principle. Then there's containers technology, like Docker, LXC, etc. Now when moving from full VM to containers the solution lightened the load on the physical host computers. Will moving to Snap packages add load back? Will there be issues when running in Containers?
With a more common operating environment for Span applications, will this not also mean more common grounds when new exploits are discovered? Are Snap packages as easily up-gradable?
How will Canonical's Livepatch integrate with Snap packages?
On computers I've gone with Snap packages, maybe this is unrelated, maybe not, but when shutting the computer down there are now several "Stop tasks", some indicating up to 10 minutes to run. In case of a power outage, if my battery in my UPS isn't spanking new, and if I haven't set my computer to stop within a couple of minutes of power outage, my machine's going to crash ! ! ! Excuse the technical language but that's NFG.
Of course I might be whipping up a storm in a teacup, but has everything been thought through? One of the reasons for going to Linux, was stability, quality and rigor. Is that rigor still all present in the Snap Package concept?
In a recent episode (Ep. 130) of the "Destination Linux" podcast the subject of Canonical and snaps was discussed and some interesting information regarding the effort required from programmers to get their software "published" to distributions using the traditional packaging formats was presented.
To those of you who do not know how a program packaged as a .deb is made available in Ubuntu (and its flavours, mine's Mate) I can recommend this 11 minute long section of the podcast, where Michael Tunnell in particular explains much of the behind the scenes work that has to be done.
The section begins at 1 hour, 2 minutes, 10 seconds
Once again many thanks to the Destination Linux team for a very useful and entertaining podcast.
Kicking this up again, since there is an official announcement regarding this in the next LTS supported version of the parent distro. I'd like to know what is going to be happening with Ubuntu-Mate as a result.
