Should wiki pages be represented as git commits or database records?

[jdmarshall]

Keeping a split source of truth or system of record would be very awkward.

One of the notable problems I encountered with Trac was that there was no AST that represented a page, so for instance identifying the hyperlinks was something that needed to be done fresh on each load, and then repeated by any plugins. So none of the plugins supported every form of link and wiki word allowed.

So from this standpoint storing the DOM into a database for later perusal would be better. It also provides a way to handle page renames, since move operations could link to a page ID determined when the link was created.

Or it could be split. Every commit has a hash, and we could build or rebuild state from that if it is missing.

Either way we still need to be able to read and scan code commits anyway, so we won’t completely avoid git integration.

But reading for would make bootstrapping a bit easier. All things to consider.