Why nobody sells an office suite any more

Ask the question directly and it sounds rude: why would anyone buy an office suite in 2026? Then look at who is actually selling one, and the rudeness turns out to be the answer.

That is not five companies being generous. It is a market that decided, over about twenty years, that a word processor is not a thing you pay for on its own. The honest answer to "why should someone buy our office suite" is that they should not, and pretending otherwise would have been a business plan built on a price nobody has been able to charge since roughly 2006.

So Rutba Office is free. Nine applications in one download, under the AGPL, no account, no telemetry, no seat count. Use it at work, open client files in it, bill for the result, and never mention us. That is not a loss leader with a catch buried in clause nine — it is what the thing is worth.

Then why build one

The unglamorous reason first: we needed the engine anyway. A business platform that issues invoices, statements, payslips and reports has to read and write .docx, .xlsx and .pptx whether or not anybody ever ships a desktop suite. That code was getting written either way, and the only real decision was whether it stayed private.

The interesting reason is that an engine is a category people genuinely do buy. Document components are sold per developer seat by real businesses with real customers — Apryse, Nutrient, Aspose, Syncfusion. Nobody is paying for a word processor; plenty of people are paying for the ability to put one inside their own product. Those are different markets that happen to share a file format.

You cannot demonstrate a document engine in a slide. You can only open somebody’s real files in front of them and let them look.

Which is exactly what the suite is for. It is not the product — it is the proof, and it is a harsher proof than a test suite, because the failures are visible to anyone who has the file.

The part that is actually hard

A .pst is not a folder of emails. It is a single-file database with its own allocation scheme and its own B-trees, and Microsoft’s specification for it runs to a couple of hundred pages. That is why so little software outside Outlook can read one, and why "just export it first" is advice that assumes you still have a working Outlook to export from.

A .docx is a zip of XML parts, and the difference between a converter and an engine is what happens to the parts you did not touch. Charts, pivot caches, macros, digital signatures, embedded media: our engine rewrites only what it deliberately edited and returns the rest byte-for-byte. Get that wrong and you have quietly damaged somebody’s file, and they find out months later, in front of someone else.

What the licence is actually doing

Dual licensing gets described as a trick, and it is easier than that. The AGPL covers everybody. A separate commercial licence covers the narrow set of things the AGPL will not allow.

• Using it — a person, a business, a school, a government department: free, permanently, nothing to register, no obligations of any kind. Copyleft binds distribution, not use.
• Embedding the engine in a closed product you sell: a commercial licence, because the AGPL would otherwise require you to publish your own source.
• Running a modified version as a hosted service: the same, because of section 13 — the clause that makes the AGPL different from the GPL, and the reason it is the licence chosen here.

And the answer for everyone who is not embedding anything

The licence answers a question most readers do not have. If you are not putting our engine inside a product you sell, the argument above is interesting and sells you nothing — so here is the other half.

We have built what most people need, and we are unusually specific about where we have not: the gaps are published in the repository, not discovered after you commit to it. Where you still see one, it can be closed at speed, and that is a thing you can buy.

• Priority support — a commitment on response rather than a place in a queue, and a route that is not a public issue tracker. That last part matters more than it sounds: the file that reproduces a fault is often one you cannot put on the internet.
• Work to order — a format read, a gap closed, an integration built, to a date you agree. We wrote all sixteen engine packages, so this is a change to the engine itself rather than a ticket passed to somebody else’s vendor.

That is the part free software usually gets wrong in the other direction — either there is no way to pay for certainty at all, or the free version is quietly the demo. Neither is true here, and both would be easy to fall into.

The same decision we made about the platform

This is the second time we have given away the obvious product, and the reasoning was identical both times: what people pay for is never the artefact, it is the burden they would rather not carry. Nobody pays for a business platform — they pay for it being run. Nobody pays for an office suite — they pay for the right to build on the engine without opening their own source.

It also imposes a discipline worth having. A customer can take the code and leave, which means everything we charge for has to be worth paying for on its own merits, on a day when they are annoyed with us.