4

Until 2017 almost all questions were to do with Excel/VBA, as described in the tag info. Since then, of the 35 questions with the tag, 28 instead refer to the machine learning compiler. There is the tag , with 81 questions, but that naming seems outdated now that XLA has moved out of TensorFlow into its own organization OpenXLA.

How about we reassign to be for the ML compiler, alias to , and create a new tag for Excel/VBA questions?

EDIT: I have added in line with the current most upvoted answer. I'd personally like a moderator to change to , but I don't understand the process.

8
  • 1
    There are 72 questions with the tag and 40 of them are about Excel addons. Where did you get those numbers from?
    – oguz ismail
    Commented Mar 23 at 3:05
  • 10
    And no, if the short tag is ambiguous no one gets to keep it. We should split it into excel-xla and openxla.
    – oguz ismail
    Commented Mar 23 at 3:08
  • @oguzismail i sorted by newest, then i split them by pre- and post-2017, as stated in the question
    – joel
    Commented Mar 23 at 3:20
  • @oguzismail re ambiguous tag, ok, though openxla might not be suitable as that's the umbrella organization, not the tech, and includes other tech such as stablehlo
    – joel
    Commented Mar 23 at 3:21
  • Your "EDIT" part does not look appropriate here. Like on the main site, on meta a question should NOT contain a verdict from the answers. If you agree with the answer, then upvote it. If you want to tell something in addition to the answer, then do that in the comments or write a separate answer. Also, stating that consensus is achieved after just a day is not quite correct.
    – Tsyvarev
    Commented Mar 24 at 18:33
  • @Tsyvarev ok sure. It's been pretty unclear to me how all this works - I have been doing my best to follow various pieces of information scattered around SO - it's not surprising I'm making mistakes. For example, how do I know when concensus is achieved? That's not explained anywhere
    – joel
    Commented Mar 24 at 19:58
  • @Tsyvarev see in particular this comment and its reply
    – joel
    Commented Mar 24 at 20:04
  • "how do I know when concensus is achieved? - Wait a day or two (weekend does not count!) and check for score on the answer. Ideally, a "consensus" is about score above 10 and without many downvotes. For "tiny" tags a consensus could be expressed in lesser score, but you still need to wait. Also, if you ask for something which requires moderator intervention, a moderator is able to determine a "consensus" by themselves.
    – Tsyvarev
    Commented Mar 24 at 22:54

1 Answer 1

Reset to default
7

We had two threads on this topic just a few weeks ago.

Aside from what you mentioned, "xla" could equally well refer to Xbox Live Arcade. Tagging a question that way probably isn't helpful, but I can definitely imagine game developers with a programming question wanting to do so.

We should not keep tags this short unless they are really definitely unambiguous - in particular, programming language names like c, c#, c++, css (I'm going to call that a programming language for these purposes, because there's a set of rules that constrain the text a programmer writes, and it's interpreted in accordance with those rules to do something interesting).

For the Excel questions, we could use a name like , sure, but I think is clearer and avoids redundancy. For TensorFlow stuff, the existing is fine, and follows established conventions.

The purpose of tags is to categorize for filtering - not to provide search keywords. If it's ambiguous, it isn't effectively categorizing.

If "XLA" referred to a compiler that weren't associated with TensorFlow or something else with similar categorizing power, we might use something like . And yes: IMO, if we ever discover something new that's worth asking programming questions about and is reasonably called "GCC", then existing questions should move to . ( just seems way too awkward.)

5
  • Sorry about the flurry of quick edits. I had difficulty getting the point across, and didn't initially understand that "the machine learning compiler" referred to the same thing as "TensorFlow XLA".
    – Karl Knechtel
    Commented Mar 23 at 3:36
  • xla-compiler's good - could move all (ML) XLA posts to that
    – joel
    Commented Mar 23 at 3:52
  • 2
    I believe XLA has been factored out of tensorflow precisely for the reason of making it agnostic to the tensorflow/pytorch/jax/whatever frontend, so I really doubt tensorflow-xla is appropriate any more
    – joel
    Commented Mar 23 at 3:54
  • also how do we come to a conclusion here and act on it? I don't know the process in meta
    – joel
    Commented Mar 23 at 3:55
  • @joel for the most part there are not really formal processes for this. In large part, if neither staff nor moderator powers are actually required to implement a proposal, people just act when they feel there is strong enough consensus - for tagging issues, that largely depends on how many questions will be affected. For example, on my own question linked at the top of my answer here, I ended up making the change I had in mind by myself, and then writing an answer to report on having done so.
    – Karl Knechtel
    Commented Mar 23 at 20:42

You must log in to answer this question.

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