User:Adamw
| Babel user information | ||||
|---|---|---|---|---|
|
Greetings! I'm a developer working for Wikimedia Germany, previously for Wikimedia Foundation, and before that as a volunteer.
This page has some of my thoughts about MediaWiki. Other ideas about the movement are collected on my Metawiki user page.
IRC: @awight
email: spam+wiki
ludd.net
pronouns: they/he
(It's already possible to read and write wikis without an internet connection, for discussion of the software see Offline_Projects.)
I'd like to see offline editorship treated as a strategic goal. Let's explore the overlaps between offline editorship, real-time collaboration, federated wikis, and sustainability. They all require a similar element of decoupling between an edit and our servers, all are goals in their own right, and they enable new dimensions of diversity.
Disinformation and other frontiers in machine learning
[edit]The could benefit enormously if we invested more in systems for automated tagging and transformation. I worked on the existing ORES system, on what is now the Machine Learning team, and see some of its limitations and potential. The new team has prepared a generalized platform for machine learning, and our challenge is now to take advantage of that opportunity.
It's become possible to identify disinformation, paid promotional material, sock puppeteering, and some abuse—we should build some of these models in partnership with the communities, and integrate into the software.
Component templates
[edit]TBD: currently writing slides for Wikimania 2021 and will post here. Templates could be more portable if their specialized dependencies (ie subtemplates and custom Lua modules) were brought inline like we've already started to do with TemplateData and TemplateStyles. Templates could be simplified if the dependency between wikitext and logic were reversed.
Other interests
[edit]- Submissions/Collaborative_drafts, a discussion I'd like to have with editors about making "Undo" more friendly.
- See a human-readable description in IdeaLab.
- Extension:Nonlinear for branching article revisioning
- Would decouple resolving any nontrivial three-way merge from making the edit itself.
- Represent revisions as a delta. Allows better asynchronous editing.
- The new visual editor will likely batch operational transformations, there's yer delta.
- Deltas would have N=||cross product|| possible combined resolutions, depending on a choice of selection. Article state at any time could be ambiguous, and multiple.
- Extension:Protection: article protection is too difficult to improve, because it is currently managed by MW core code. Once these permissions have been componentized, it will be more clear how to hook in alternative permission schemes.