The problem is that re-defining definitions brings in chaos and inconsitency in science and publications.

Redefining what a "planet" (science) is or a "line" (mathematics) may be useful but after such a speech act creates ambiguity for each mention of either term -- namely, whether the old or new definition was meant.

Additionally, different people use their own personal definition for things, each contradicting with each other.

A better way would be to use concept identifiers made up of the actual words followed by a numeric ID that indicates author and definition version number, and re-definitions would lead to only those being in use from that point in time onwards ("moon-9634", "planet-349", "line-0", "triangle-23"). Versioning is a good thing, and disambiguating words that name different concepts via precise notation is also a good thing where that matters (e.g., in the sciences).

A first approach in that direction is WordNet, but outside of science (people tried to disentangle different senses of the same words and assign unique numbers to each).