May 10th, 2012

beartato phd

(no subject)

Here's someone's blog post about a category-theoretic thread of thinking for resolving two edits to the same text, a la version control systems, that is new to me. What's interesting about it is that it describes a category that has all pushouts, which goes against the intuition that not all pairs of edits have a sensible resolution. It resolves this dissonance by admitting into the category more objects than you might have expected, including ones with "two parallel tracks" at the site of the conflicting edits, thereby postponing the conflict resolution step to a human decision of how to linearize the conflict-state object. I don't fully see how this works out, but I rather like the sound of it.