What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 128 Stack Exchange communities.

When will [on-hold] questions actually be migrated?

I've seeing lots of reasonable questions being put "on hold", just because someone deemed some other SE site more appropriate. Questions are being disabled, but never actually migrated and reenabled?

E.g. these questions, which are reasonable although not "perfect":

At the same time, SO doesn't really help users in choosing the most appropriate site, or migrating their question:

Furthermore, we are seeing more and more and more sites pop up, often with a limited active community; so that it may be sensible to ask on an off-topic site, because there are actually more on-topic active users there...

E.g. to pick on the first example above. It's not a great question (he didn't share what he tried to compute, so we could improve on his code). But where does it really belong?

  • StackOverflow, because it is a programming question?
  • Cross Validated (Stats.SE) because it is a question about variance?
  • Machine Learning SE (in Area51) because some people consider k-means to be a "learning" algorithm?
  • Computer Science stackoverflow (in beta), because it might be computer science (or statistics ... but probably not "theoretical computer science", which is yet another site: cstheory.SE!)
  • Math.SE or MathOverflow (yes, there are two), because it definitely involves math...
  • Maybe a "data mining" site, that focuses on unsupervised methods (i.e. "not learning")

I guess on each of these sites you will find k-means questions... over 300 on SO and e.g. over 100 on stats. just closing them sometimes, migrating them sometimes, does not do the community a favor IMHO.

In my opinion, we should try to:

  1. Get a much clearer definition of what is on-topic for which site, as they overlap quite a bit
  2. Make migrations faster, without disabling questions for days
  3. Help users much more to choose the right site the first place. Maybe even with simple rules or tags; e.g. "server cluster -> suggest superuser", "k-means -> suggest crossvalidated", "complexity -> suggest cstheory"
  4. Maybe reduce the number of sites overall on the long run, it breaks communities

Or maybe we can somehow have the sites just be merged back together. And associate tags with sites. I.e. the tag "kmeans" is associated to the "crossvalidated" view, whereas "complexity" is associated with the cstheory view. Right now, you always have to check a dozen sites...

share|improve this question
If there is a consensus in the close votes as to where a question belongs, it gets migrated automatically (4 out of the 5 votes). When this doesn't happen to a question, it tends to indicate that it is probably not a good fit for migration. –  Oded Jul 6 '13 at 11:52
As for the specific questions - The first is not a programming issue it is (possibly) about algorithm choice. The second is very low quality and also not a programming question. The third is asking too many things in one question. All three are examples of questions that should be [on hold]. –  Oded Jul 6 '13 at 11:54
@Oded there won't be a consensus when there are only 5 sites to choose from (not including Cross Validated, for example). Which is probably why I'm seeing this happen so often. –  Anony-Mousse Jul 6 '13 at 12:28
Heh. You thing that if you had all +100 sites available you would get a consensus? Chances are better that you won't... –  Oded Jul 6 '13 at 12:30
Well, chances are that already the first one picks crossvalidated for k-means questions, and the others then agree... You could easily place the already-picked ones on top of the list. Or use a primitive recommendation system to offer a choice of 5 relevant ones, instead of always giving the same choice. Chances are that 95% of the sites are obviously not relevant for the question. –  Anony-Mousse Jul 6 '13 at 13:20
The kmeans Q looks like it might be a stats Q, but it's really more likely they know "how" mathematically they just haven't shown what they've tried yet in terms of code. The time series Q is definitely a stats Q but it is far too low quality to migrate. I have no opinion on the third. –  joran Jul 6 '13 at 13:29
Note that I don't care much about the actual questions, but about the overall process of migrating questions. It doesn't seem to work well for most sites (probably all but the ones in the shortlist, where you can actually vote for a particular site). And judging e.g. from meta.stackexchange.com/questions/12570/… we really need to help users to post to the right site... –  Anony-Mousse Jul 6 '13 at 14:05
add comment

You must log in to answer this question.

Browse other questions tagged .