I worked in this space for a loooooong time (mostly in higher ed; I have less knowledge about the corporate or k-12 segments.)
The short answer is: It's surprisingly complex to build a good LMS at all; it verges on impossible to build a good generic LMS that covers all or even many use cases.
Everyone teaches differently; everyone has their own particular method for grading; everyone has their own opinions about how much or how little classroom communication should be handled through the LMS vs in person; everyone has their own preexisting tools and data they need to integrate with. All of those differences conflict with one another to some extent; class sizes differ, class durations differ, etc etc etc. Supporting all of these would guarantee an overcomplicated and difficult to use interface.
This isn't the sort of thing that open source tends to do well. The best open-source work is "scratch your own itch" kind of stuff; this pattern in LMS development has historically been that they start out that way, as a home-rolled solution built by a tech-savvy instructor, then accrete layers of contradictory features in pursuit of a larger audience, and either become too unwieldy to use in their own right or else get purchased by one of the already-bloated large competitors.
(I find it personally hilarious that you cite Canvas as the "archaic UI and features" state of the art: Canvas found its original success as the clean, modern alternative to BlackBoard, which at the time was the bloated-to-the-point-of-unusability standard tool. The cycle continues.)
There are also a lot of regulatory hurdles around accessibility, privacy, and pedagogical standards -- particularly in K-12 ed -- that, while very necessary and useful, take a lot of work and knowledge to accomplish, and serve as a barrier to entry to new products in the space.
The short answer is: It's surprisingly complex to build a good LMS at all; it verges on impossible to build a good generic LMS that covers all or even many use cases.
Everyone teaches differently; everyone has their own particular method for grading; everyone has their own opinions about how much or how little classroom communication should be handled through the LMS vs in person; everyone has their own preexisting tools and data they need to integrate with. All of those differences conflict with one another to some extent; class sizes differ, class durations differ, etc etc etc. Supporting all of these would guarantee an overcomplicated and difficult to use interface.
This isn't the sort of thing that open source tends to do well. The best open-source work is "scratch your own itch" kind of stuff; this pattern in LMS development has historically been that they start out that way, as a home-rolled solution built by a tech-savvy instructor, then accrete layers of contradictory features in pursuit of a larger audience, and either become too unwieldy to use in their own right or else get purchased by one of the already-bloated large competitors.
(I find it personally hilarious that you cite Canvas as the "archaic UI and features" state of the art: Canvas found its original success as the clean, modern alternative to BlackBoard, which at the time was the bloated-to-the-point-of-unusability standard tool. The cycle continues.)
There are also a lot of regulatory hurdles around accessibility, privacy, and pedagogical standards -- particularly in K-12 ed -- that, while very necessary and useful, take a lot of work and knowledge to accomplish, and serve as a barrier to entry to new products in the space.