I noticed that both meshing & mesh-generation are included as tags. Should we merge them or at least suggest a tag-synonym?
-
$\begingroup$ Not sure. I'd like to hear what the PDE guys have to say (i.e., I'll prod Jed, David, et al). $\endgroup$– Geoff OxberryMar 4, 2012 at 16:57
-
$\begingroup$ Perhaps we want to distinguish "mesh selection" from "mesh generation?" (I'd argue "meshing" includes "mesh generation," but "mesh selection" would be potentially an independent topic.) $\endgroup$– aeismailMar 4, 2012 at 17:36
-
$\begingroup$ @aeismail: what do you mean by "mesh selection"? $\endgroup$– Paul ModMar 4, 2012 at 17:38
-
$\begingroup$ @Paul: What I wanted to distinguish was theory from implementation. Saying how a mesh should be created versus the actual creation of the "optimal" mesh. $\endgroup$– aeismailMar 4, 2012 at 17:41
2 Answers
I think the best approach is to introduce synonyms. Tags serve as a classification as well as a keyword for search. Depending on the scientific community, one term may be more prevalent than another. Merging diminishes the number of false negatives in search of specific topics.
I agree with Deathbreath that synonyms are the best approach. My philosophy on tags is as follows:
- Fewer is better. Tags are supposed to represent a level of abstraction and have some leeway in them. I'd rather have sparse-linear-algebra (or some tag like that) be an umbrella tag for sparse linear algebra methods than have a lot of singleton tags for every single solver (i.e., sor, qmr, gmres, etc.).
- Spell out acronyms unless unreasonably long (or unless they're software packages, e.g. vasp, petsc). I'd rather have large-eddy-simulation and discontinuous-galerkin, than les and dg, respectively. The former tags are more Googleable and descriptive than the latter. Even if tag wikis can mitigate the crypticness of acronym tags, not all tags have blurbs accompanying them; Paul deserves to be recognized for doing the lion's share of the work writing tag descriptions.
- New tags should only be created if the question can't possibly be described by existing tags (which is why I created interval-arithmetic recently) or there are so many questions (at least 5 to 10) within an existing tag that can be grouped into a useful subcategory, in which case a tag should be created for the subcategory (i.e., if we get enough questions about GMRES specifically, then we should create the tag gmres; otherwise, it should go under something like iterative-methods).
Tags are supposed to be a nice way of categorizing and indexing questions, and you can tag a question in at most 5 categories for a reason: more tags per question would likely be unwieldy. Having too many tags makes it more difficult to search through tags on the "Tags" page, and it then requires you to search by typing in the "right" category in the search box. Unless it occurs to you to search for qmr or minres, you're unlikely to find those categories by browsing the tags because they appear at the very end of the list on the "Tags" page -- they are each associated with one question. Having not enough tags means that you have to filter through a lot of questions in a single category to find what you want. We should strive for a happy medium.