It seems like everyone has had their word on the latest release, but like a fashionably-late party-goer, I’m going to waltz in at the 11th hour and offer my 2p 🙂
I think it’s well-known at this point that 14 has shaped up to be a very good release, but I’d like to draw attention to one point in particular: the version of Nouveau in this release is another big, big step forward. I have a relatively bog-standard Dell D830, and 14 is the first time that:
- suspend/resume has worked out of the box – this is huge for me
- the Mesa 3D drivers, although marked experimental, work well enough to run Compiz easily
Is nouveau’s performance great? No, to be honest, it actually feels slightly slower here than on 13 (although almost certainly because I’m now in Compiz, not metacity) – but for me, this doesn’t matter, being able to suspend is massive. I could even envisage the 3D stuff being turned on by default in the next release or two.
If there’s part of the system which sticks out as still being sub-optimal, though, it’s the application install experience. I know I’m not saying anything new here, or probably anything anyone disagrees with. A great example is attempting to install OpenOffice.org on a clean Live install (OOo no longer comes by default on the CD), because you have to negotiate a couple of problems:
- you have to figure out where the openoffice packages are (searching on “openoffice” isn’t enough sadly; it pulls through large amounts of non-openoffice packages);
- once you’ve found the packages, you have to figure out which ones you need – amongst a sea of langpacks, extensions and other stuff, are the bits you actually need. Calc is relatively easy to find; Impress less so, Writer comes right at the bottom (alphabetical, see) – not easy. Plus then there are the bits you do actually want – extra graphics filters, extended PDF support, etc.
- then, when you’ve figured out which bits to install, you set it going and the “success” dialog looks an awful lot like “fail”:
[<img class="aligncenter size-full wp-image-382" title="OpenOffice.org install fail" src="http://www.alexhudson.com/wp-content/uploads/2010/11/openoffice-fail.png" alt="PackageKit dialog coming up with no actual content." width="394" height="214" srcset="https://www.alexhudson.com/wp-content/uploads/2010/11/openoffice-fail.png 657w, https://www.alexhudson.com/wp-content/uploads/2010/11/openoffice-fail-300x162.png 300w" sizes="(max-width: 394px) 100vw, 394px" />]They are lovely icons, though 🙂
I’ve previously said that I don’t really understand why all of these types of installation tasks are grouped together in the same application: for example, my belief is that font installation is much better served by a Google Fonts-alike web service which can be used to browse and try fonts live: you’d then hit an “Install” button or something which would then trigger the excellent PackageKit. However, many people remain unconvinced even in the face of the actual numbers and things.
It’s the same with this. We need some kind of application store. I personally don’t see why this should be conflated with the package mechanism: packages are the how, the app store is the what. It makes no sense, to me, that all packages be treated identically: for example, if an application can talk to the PackageKit interface for its plugins/extensions, there’s no reason to have that stuff in the app store at all. Similarly, just because something isn’t a GUI application doesn’t mean it should be excluded: why can’t we have a “Python Developers’ Corner” in the store to browse libraries and things? That’s what I want as a user (yes, developers are users too).
This isn’t going to get fixed quickly, and sadly I think efforts like Ubuntu’s Application Store don’t solve many of the problems: if your application store is just a majorly cut-down view of the package database, I think you’re doing it wrong (for one thing, it doesn’t scale as you add back in all the packages you cut out).
At some point over Christmas I might have a go at attacking this problem; a lot of the pieces needed are already in place: PackageKit is more than capable enough of installing things from the web, and I really think that having an actual prototype that people could use would do so much to illustrate the idea that even if it wasn’t used, it would help push things in a better direction.