I've been improving/organizing tags since some time now, and encountered a few cases where I don't want to advance without any feedback from the community. Below I suggest some changes/merges that I find useful, please provide feedback. For sake of simplicity I won't create a new post for each suggestion, as the list is big. I've provided some explanation of my intentions at the end of this post.
As a matter of good practice I also would like to suggest the followings for tagging:
- Rather combine existing tags cleverly than introducing new ones.
- If you insist on introducing a new tag please add a wiki as well that helps to define the domain that the tag is intended to cover.
- Please do some research before inventing new tags to prevent parallel concepts to arise (like matrix and linear-algebra).
I've added each suggestion as a separate answer, and I encourage you to do so in the future to. This way we can canalize occasional retagging events to this thread.
Naturally such categorization processes as tagging involve a lot of subjectivity, so there are always tags that are important for a few people only, or tags that were introduced by different people for identical (or only slightly diverging) concepts, which results in more tags than necessary. I would strongly advice the combination of existing tags when posting a question instead of introducing new ones, as that would lead to a combinatorial explosion. A smaller lexicon of tags if combined leads to higher generality (i.e. can cover a larger domain) than a set of highly specialized tags (of course these are only the two extrema). At present we have a fair amount of tags that reasonably covers all the questions asked so far, and this amount is still maintainable.
Why not simply keep all tags? Why not have millions of tags? One thing is that a larger lexicon is less general (specific tags cover specific domains), therefore less useable by the community (specific tags would be used or understood possibly by very few people) and less maintainable (no one would be able to scan through a vast list of tags for e.g. relationships). It also could have detrimental effects, like supporting behavior that is unwanted (i.e. an existing code-request tag might give the wrong impression that we support "gimme-teh-codez" type questions, see here).