Time  Nick    Message
15:44 mikemol I expect to need to default to method 2, because the database/disk intensity of method 1 will be very ugly. User preference is needed to allow the user to switch to defaulting to method 1.
15:43 mikemol (And the render-on-request does the check-and-cache behavior in the database)
15:42 mikemol In method 2, we default to not sending highlighted code to the client, and we default to having the relevant div collapsed. A user-initiated javascript event queries the server for the highlighted form of the snippet.
15:38 mikemol Er. Call that 'method 1'. 'method 2' is what I was thinking about this morning, and differs from method 1 slightly.
15:38 mikemol When a page is updated, purge the database of all code snippet render cache entries for that page.
15:36 mikemol If a current-version page is being rendered, check the database for rendered version of the snippet as rendered with the selected rendering options. If it's there, use it. If it's not there, render, commit rendered version.
15:31 mikemol 4) It's going to need its own table in the database.
15:30 mikemol 3) I should probably stop supporting NoScript.
15:30 mikemol 2) It's going to need a setting in user preferences to tune some of its behavior.
15:29 mikemol 1) It's going to need to be aware of the difference between pages in edit 'preview' mode and pages which are committed (and are the current revision).
15:28 mikemol I realized what I should do for the MWGeSHiGlue extension, but it's going to be a lot more complicated than what we currently have.