- My big ol' pile of facts system is flexible, but accumulates a lot of stuff that's just referenced or mentioned by me. Distinguishing the canonical records from those items from the canonical records of items I actually created is important, particularly once I actually generate information about projects, etc.
- Loads of old LJ posts used table based layouts to format a mosaic of photos. Translating those to use CSS grids or something along those lines would be a nice upgrade.
- Tumblr id
459649779has a raw YouTube embed in the body that thetoMarkdown()function is stripping out rather than keeping as a free-floating URL. - Tumblr Link posts often have 'via X…' metadata about the original source of the share. The actual share links are in some weird places, though, and deserve some extra attention. Right now, they're unhandled.
- Pinboard has copies of my Delicious links; deduplicating them and preserving the Delicious attribution would be nice.
- Clean generation of 'threaded sort-order' keyes for arbitrary comments is now in place. Woo. It duplicates Drupal's comment-sort ID generator, and works on imported Disqus threads as well. I'll be using it on Twitter threads as well, to preserve hierarchy on threads with weird discursive sub-trees of interspersed replies.
- Right now it's hard coded to use
id,parent, andthreadproperties: that works with the CommentSchema entity but for other data it'd be nice to make it configurable.
- Right now it's hard coded to use
- Split the schema stuff into a secondary project that can be imported into 11ty etc easier
- Good lord, the saveThimg() and linkThing() kludges are terrible. Special casing each backend is madness. Ultimately I'd like to support KVS on the local filesystem, ArangoDB, Postgres, and Sqlite. That shouldn't be TOO bad, but will take some discipline, particularly around the way that IDs are created and exchanged. settling on URNs (
urn:fyi:type:idwould probably make sense, and internally could be shortened totype:id)