It would seem to me that bundles of bundles would be superior to bundles of packages. I find myself wishing that I could install a subset of a current bundle. An example would be SDDM. It is only bundled in desktop-kde. I want the DM not the desktop. If it were a bundle, I could install just the DM.
Why does a “package” exist, wouldn’t a bundle of that package also work?
I have similar issues. There are some subset of programs I think should be offered in individual bundles. For example I use tiling window manager, in most Linux distros I can download each of the parts I need separately (GVFS, mate-polkit, PCManFM, etc) and build a fully functioning desktop environment.
With the current bundle system I would need to download all of gnome, lxqt, and if they offered it, mate desktop, to get the results I needed.
While clear Linux does offer devpkg, not all the programs are available and there isn’t an easy way to get the individual devpkg needed without grabbing a whole bunch of them you don’t want (currently).
You might want to learn how to use mixer tool, in theory you should be able to use a combination of devpkg and ‘exclude’ rules to mix your environment of choice.
I looked into the mixer option previously in my exploration of CL. However, to create a custom version of CL requires one to maintain a swupd server and that seems more than we care to do. And to build a mix, you add “bundles”, much the same issue. If I want to build my OS, I would go with Gentoo, which is designed to be built. I have found configuring CL for a production work environment to be challenge enough.
When we started Clear Linux, we had a few bundles which we tried to cater to broad use cases. Feedback from users indicated that a more flexible solution was needed and we have been increasing the number of bundles to improve the granularity of packaging. As always with these things, there is a balance between flexibility and ease of use with the demands of bundling every subset of programs individually. If there are are specific use cases where we could improve, please post an issue in Github with all the details and examples, and we can have an engineer review it.
Rgds,
Chris
If you agree that cost of disk space is negligible comparing to the time and effort spent on managing the system as an administrator, it really makes sense to have bundles that install tons of packages instead of installing them manually.
Bundles are like a higher level abstraction to packages. Normally you start with a need to accomplish something, you find a package that meets you needs, and then you install it, this is unnatural. The natural way is, I want the system being capable of dealing with certain types of tasks, and if I need a specific functionality, it should be readily available.
That is too say, bundles do come with things you don’t need at this point, but they’re just lying there and when you need something, after necessary configurations it can be deployed right away.
Functionality and Convenience over disk usage efficiency.
Another pro is, that the more bundles you install, the faster they install, because more dependencies were already installed before, with other bundles.
I generally agree with the points about installing additional software. However; in this instance, I am configuring CL for a secure production environment. We can not permit people to have access to command lines and the likes of wireshark (unless their job requires it). We even need to prevent the deleting of some files in their own /home directory (thus no gnome3 or KDE5 with their integrated file managers). We try to limit the amount of software on the clients to that which is required to perform the job, less patching and fewer security concerns… without being too Orwellian about it.
@0n0w1c - it sounds like for that level of control, you should look into owning your own mix of Clear Linux. That way you can control exactly which bundles would be available to your users from your own swupd server.
It may be that I eventually end up there, the thought has crossed my mind multiple times. However, since I am responsible for a production environment, I need to be pretty sure of both the short term and the long term. What would be nice is to have a “Life with a swupd server” manual. While I can read about making a mix… how to live with the result, what can and will go wrong, etc.
I took a strong enough forward stance recommending CL for our environment. My hands are still full simply configuring CL to operate in our RHEL shop. I do not have a full command of CL and for me to adopt a “I’ll just build it better” approach is probably foolish at this time. Given more time, I’ll probably get there. 
And to be fair, I don’t know of anyone using a mix in production yet. There may be some bugs in there that you don’t want to be the first one to find Having said that, if you did want to go down that road, I’m sure we’d be interested in working with you to wring any issues out.