Open source is not licence free

The most expensive sentence in an open-source migration is "it's free, so we don't need to think about licensing." Open-source licences are contracts. They grant broad rights and attach real obligations, and the obligations differ sharply between a permissive MIT component and an AGPL application at the heart of your stack.

Copyleft is the one that surprises people. GPL and AGPL conditions on distribution and network use are exactly what keeps these projects alive — but if you modify an AGPL system and expose it to users, you have obligations most procurement checklists never mention. None of this is a reason to avoid copyleft software; it is a reason to know, before deployment, which licence sits under every system you run.

Then there is the commercial layer on top: open-core products where the community edition is genuinely open but the features you eventually need are not, dual-licensed projects where the terms depend on how you use them, and vendors who relicense between versions. A licence review is not paranoia — recent history is full of teams who woke up on a different licence than the one they adopted.

The honest comparison is total cost of ownership. Open source moves spend from per-seat licence fees to implementation, integration, and operations — money that builds capability inside your organisation instead of renewing outside it. And it removes the worst line item of all: the renewal letter whose price you cannot negotiate because leaving would hurt more.

Our rule for clients is simple: every system in production gets a licence entry — what it is, what it obliges, what the commercial edition adds, and what an exit would take. It is one page per tool, it costs an afternoon, and it is the difference between owning your stack and merely not paying for it yet.