This question just made me look at the tag , and I didn't like what I saw. The tag is a mismatched collection of questions about

  • Lorentz covariance of theories in relativity,
  • covariance of a physical quantity under Lorentz transformations,
  • covariant derivatives,
  • the differences between covariant and contravariant vectors, and
  • statistical covariance.

For a tag to be useful it needs to refer to a coherent set of connected questions; by that metric this tag is useless. I therefore propose we instate the tags

While statistical covariance has huge traction in physics (e.g. this), this is our only statistics covariance question and it's hardly about that, so I propose that tag be dropped. If needed further down, the tag can be introduced.

I will implement these changes over the weekend (so I don't dump forty questions on the front page in one go) unless there are any objections or better proposals.

To help prevent this happening again, I also propose we blacklist the tag. Setting up as a synonym for will ensure it will come up when someone tries to use , and having appropriate tag wiki excerpts for the tags should help guide people to the correct tag.


1 Answer 1


I) Specifically about covariance:

1) The tag is supposed to cover whatever-notion-the-physical-system-in-the-question-is-covariant-under, typically relativistic covariance, but also more abstractly, e.g., contravariant and covariant tensors not necessarily in the context of relativity.

All other uses of the tag than the above are wrong.

2) Covariant derivatives in differential geometry should be tagged or and perhaps . (The tag could apply here if the main topic of the question is point 1.)

3) Gauge Covariant derivatives should be tagged and perhaps and .

4) Covector, bivector, differential forms, etc, should typically be tagged with or .

5) Statistical covariance, average, etc, should be tagged with (which in turn should only be used for statistics in the mathematical sense. E.g. Fermi-Dirac statistics etc. go under and perhaps , while the tag is ideally reserved for issues related to the spin-statistics theorem).

II) More generally, the issue of tagging is non-trivial and difficult to make perfect. Since the system doesn't know which tags are close, and which tags are far apart in tag space (the metric is the discrete metric), it is often better to have broad tags with simple rules rather than specific tags that nobody else will use consistently or have heard about. In plain English: it might be better to compromise and either (i) tag dromedary questions with the tag rather than introducing a new tag with the consequence that the system will no longer recognize that posts with and are likely very similar; or (ii) make a synonym of the tag, with the consequence that questions tagged with would automatically be retagged as by the system, even if experts of dromedaries would strongly object to this.

There will likely continue to be tags that cause various problems. E.g. the tags and are often applied wrongly.

Non-descriptive tags, such as, e.g. , , , etc, should be avoided as they serve little purpose. (Many questions on Phys.SE involve differential equations and equations of motion.)

An ambiguous tag, such as, e.g. should either be replaced by , , , , etc, depending on context. Or the proper use of an ambiguous tag should be spelled out in the corresponding tag wiki. E.g. use of the tag is explained in its tag wiki.

And so forth.

  • $\begingroup$ Ok, that's all nice and good, but there still needs to be a way to prevent misuse of the tag. $\endgroup$ Mar 27, 2014 at 22:46
  • $\begingroup$ The logical way to do that is make covariance a synonym of something. $\endgroup$
    – David Z
    Mar 28, 2014 at 4:29
  • 2
    $\begingroup$ The problem with this one, to borrow your animal metaphors, is that we have a bunch of questions tagged eight-legged-animal that should either be tagged spider or octopus. There is no synonym of covariance which will cover all of the questions it currently has. $\endgroup$ Mar 28, 2014 at 12:45

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .