He started with several reasons of why and when not to transition:
- When there’s neither a benefit for the bottom line nor for the customers/users
- When it leaves you without a productive and operable documentation environment at any stage
- When you cannot commit to the new process (i.e., you’re afraid to burn your bridges behind you)
- When you cannot train or run a pilot project
To select what to carry over, he advocated converting only the used, required contents – and when in doubt to leave it out. He also said to carry over all the existing semantics, but not to add any new semantics, unless they add a proven value at reasonable cost: Don’t let the perfectionists run wild and build a more perfect solution – it’s difficult enough as it is.
For the “how”, he suggested to get outside help, esp. with the actual conversion – though that may require to clean up legacy content to get it ready for sensible conversion.