To be free we need usable data, so for instance instead of strange zipped file formats a spreadsheet (witch should not exists, but that's another story) should save data as *sv/SQLite DB by default, a visual doc should be LaTeX or something similar and so on. The suite should be only a viewer.
I with you on sqlite but lost me on LaTeX.
Maybe, many open formats, with many open tools?
And I'm over wishing, hoping for better print support on HTML - I think is the holy Grail.
First crap is the WYSIWYG model, it produce not so nice results and demand very complex UIs. To produce high quality documents LaTeX so far is the best tool, PostScript itself is not known essentially anymore, {La,}TeX is often wrapper for instance in R RMD/Quarto docs to produce nice printouts, it's the de facto standard of scientific typeset and have so far no better competitors, so it's a wise choice.
Second crap is the spreadsheet concept, yes sometimes we need tables, and compute something on them. Org-mode show a simple example of table computing without being a spreadsheet, R is another, most modern NotebookUIs from Mathematica to Jupyter do the same as well. Spreadsheets was an idea to offer limited data manipulation to computer illiterate, and failed. The long tail of disasters suggest it's time to abandon such model. So we need something to store data. sv works well for non-giant dataset, they are simple than some ML dialects, for more SQLite is a popular powerful tool offering a self-contained easy to move storage. I do not like it, but for various things is good, for the rest it's deadly simple to export the DB in some other formats.
We should re-create the Smalltalk desktop model, the end-user programming concept that happen anyway today but hyper limited and in user-unfriendly ways. Take a look at Pharo, than image not just the bare bone environment but something developed like Emacs today. You imaging this: http://augmentingcognition.com/assets/Kay1977.pdf or http://www.bitsavers.org/pdf/xerox/xsis/XSIS_Smalltalk_Produ... for the modern world.
Back than it fails because of high costs and general ignorance, people simply do not understand the power of this model so feel no need of it. Such legacy burden plague much the present IT as well, with people still imaging files as sheets and folders as suspended folders full of paper sheets in a drawer. Mentally most people have issue going past the paper tool even today. But today we have hw costs small enough and computers themselves spread enough to push again such model.
This model is the nightmare of all big of IT, from old IBM to modern GAFAM, because they know they can't compete with a FLOSS networked desktops model. That's why they do what they can to keep people on the old mainframe model, now the cloud, and many ignoring this keep mimicking the big e powerful not understanding that they try to replicate an archaic model born out of some Xerox spoils once they see how to change them in anti-users ways.
The "FLOSS alternative to" is the networked, end-user programmable desktop model, not some sightly different clone of some GAFAM platform.
Re WYSIWIG:
Rather than WYSIWIG being bad, I think the introduction of scroll bars destroyed WYSIWIG from the start. It only works on a large monitor where one page can be viewed and worked upon. In the early 1990s it was only at uni or in the workplace that this was possible. Not at home on a 12 inch monitor. It's the same reason that a web page that is larger than the screen is a health nightmare (mentally and for poor mouse wrists).
Re computing using spreadsheets:
This made computing available to people who couldn't program imperatively. It's given more power to the average power user than you can imagine. See Scott Hanselman's interview with Oz du Soleil. https://hanselminutes.com/532/data-literacy-and-the-usefulne...
Re data storage:
Without the ability to program in a declarative way (i.e. SQL) it is going to be difficult to both store and retrieve data in a provably ACID way, yet non-programmers still get it done well enough with files.
Re Smalltalk:
Without OO knowledge, this great tool will not be useful to any desktop computing power user that has not formally learnt programming.
My point is that the tools that have survived and got the job done in the "computerised office" have not required the understanding of a complete pseudo-math model (aka programming paradigm) in order to get useful work done. People did not need to be able to compile a program without errors. They used tools that can be iteratively hacked upon, and allowed non-purist tech people to become power users in the workplace.
TL;DR
Models of software environments that do not have to be programmable have proved themselves and the people who use them capable of getting the job done. That can't be replaced by your suggested model.
For the author and I suppose for you as well, programming means create a project, an UI, publish the result as an installable package and so on. Yes, that's not for any computer user. But that's not end-user programming as well.
As I said before my OS is Emacs/EXWM (booted by NixOS), so I have most of my digital life in org-mode, org-roam managed notes, OS config included tangled from org-mode notes. My user programming is for instance a quick script to auto-archives pdfs bill from my mobile and landline, ... in their right place managed as org-attach-ments, creating a relevant org-mode heading with a BeanCount babel block. This means that with not much effort, no specific project, UI design, packaging etc I automate a generic end user task in a way no pre-made tool can do. This is not harder than using Excel or MS Office macros, but the outcome is far, far better. With this model I have a full-text search built-in, simply via ripgrep wrapped in Emacs to get live results as I type, no special setup/indexing needed. All my automation, shell scripts, elisp snippets etc fell in this form, single listings, a file at maximum, something simple enough for a non-programmer (like me, because I have a CS degree but I'm not a programmer). Anything fit the same UI, I can link mails (notmuch/ol-notmuch), transactions (BeanCount), binary files that demand specific GUIs to be read/manipulated (a simple elisp: org-mode link) and so on.
This can be done by ANY computer user, if he/she is trained to this model instead of the Windows one. I know for personal experiments with some friends children. While the Windows model end up in the Android model and in the https://www.theverge.com/22684730/students-file-folder-direc... because that's needed for GAFAM et al. business. The classic model end up in an educated userbase who grow and learn with pleasure for their lifetime. While stuff you learn for Windows or web stuff is rubbish in few years simply because the upstream is changed.
The model I propose is the school model, we start learning, learn a bit for some years, profit from life keeping learning, producing new things. The Windows/GAFAM model is those of the ignorant user, where anyone can start clicking around and arrive to a point of being able to do something, still unable to grow, still not mastering their tools. This models seems to be quick at first, but it's a dripping project that ensure persistent frustration and ignorance, impeding innovation and controlling the users like sheep.