25

The migration of a question is automatically rejected when none of its tags exist on the target-site. I believe this feature creates more issues than it solves and thus would like to see it gone. (This is an out-growth of a discussion on the meta of MathOverflow, which I repost here as it concerns several/all sites.)

The main reason why I think so is that there can be numerous reasons other than the subject being off-topic why some tag does not exist on some site, for one thing it could be a spelling, typesetting, or other formalistic reason. For example, for the two sites I know best MathOverflow and Mathematics hardly any of the most popular tags have a strictly matching tag on Mathematics but only because on MO there is the convention that certain tags are formatted in a specific way, which is not in place on Mathematics (for example nt.number-theory vs number-theory).

On top of that a user choosing the wrong site in the first place is likely not too well aware of the tags of the site they are posting too so that the question might be poorly tagged in the first place.

There is an additional detail I do not understand (it is obsolete if the actual request should be followed but in case it is not):

Even if on the source site a tag has a synonym and that synonym exists as a tag on the target site the migration is rejected. I do not understand the logic behind this at all. Possibly this is just an oversight.

I could also think of some other ways to improve the process while keeping the idea, but since I believe that it is best dropped I do not (yet) elaborate on this.

  • 3
    This seems like something that the two math sites should hash out and potentially get an exception for just their case. The purpose of the feature is to prevent migrating questions which have no place on another site. I don't think your observations apply globally to the network. – animuson Dec 7 '13 at 0:52
  • 1
    I understood the intended purpose of the feature. Does it do a good job anywhere? (In the linked to discussion two moderators from non math sites did not seem enthusistic, to say the least, about this feature either.) – quid Dec 7 '13 at 0:56
  • 4
    I'd be interested to see a list of some of the questions which have been auto-rejected. I'm not sure how you'd go about getting something like that. Might need a staff member. – animuson Dec 7 '13 at 1:09
  • I'm with @animuson. Let's see if someone can provide some stats and I night be inclined to agree. On the surface, your proposal makes some sense, but in practice, I'm curious about what actually gets rejected for this reason. – psubsee2003 Dec 7 '13 at 1:51
  • @psubsee2003 I agree it would be interesting to have data (and I also upvoted the respective comment), but I think some things can also be approached in abstract based on related data. Looking at meta.stackexchange.com/questions/103168/… demonstrates that minor deviations in spelling are a considerable issue (even on a single site, not cross-site as is relevant here); this is for SO, I know it is the same on MO and presumably is an issue everywhere. – quid Dec 7 '13 at 14:49
  • 2
    I don’t understand the intended purpose of this feature at all. Why would anyone migrate questions which have no place on the other site? Is the assumption that all four people required to migrate a question are idiots, and therefore liable to choose a totally wrong target site? – Emil Jeřábek supports Monica Dec 7 '13 at 14:53
  • 5
    Anyway, different site have different scopes (that’s the whole reason why migrations are done in the first place), and therefore different needs how to organize the reality they are dealing with by means of tags, and different naming conventions. There is no reason to expect that the proper way to tag a particular question should be the same on both sites other than by chance, and conversely, there is no guarantee that tags that happen to have the same name actually denote the same concept. The whole check thus feels like a lottery. – Emil Jeřábek supports Monica Dec 7 '13 at 15:12
  • 5
    However, the most nasty feature of this check is that it does not give any sporting chance of preventing failure. There is no advance warning that the question will be automatically rejected, close voters are presented with the migration option even if it is in fact impossible, and when the time comes, the system silently drops the migration. The overall behaviour is as if you send a packet by a messenger, who has the habit that after he turns the first corner, he rolls a couple of dice, and if the right points turn up, embezzles the delivery. – Emil Jeřábek supports Monica Dec 7 '13 at 15:19
  • 2
    The only way to guard against this would be for someone to volunteer to manually check the tags on the target site, which is time consuming, tedious, and error prone, needs to be done for every question considered for migration, and is not remotely worth the trouble. If people are expected to do menial tasks in order to work around random technical quirks of a computer system, that’s a hallmark of bad software design. – Emil Jeřábek supports Monica Dec 7 '13 at 15:26
  • 3
    I agree @EmilJeřábek to have the issue pointed out when voting surely would help. Conversely, if one is aware of the issue but wants to work around it one could just create the tag on the fly when voting. Which IMO is further illustartion that this is really just something that can bite you when being unaware but not providing 'protection' against much anything.// Regarding examples not exclusive to math. On MO we hava a tag latex, when used chances are the q is for tex.SE. Yet while they have twenty-four(!) tags containing latex no tag just latex there, so it is off-topic there. – quid Dec 7 '13 at 15:40
  • @shog9 now that it's declined no point in the bounty, unless there's still a chance to reconsider? – Shadow The Princess Wizard Jan 13 '14 at 8:30
  • The bounty just states he wants it to have more attention; maybe Mad just wants more folks to be aware of this behavior, @Sha? Who knows. – Shog9 Jan 13 '14 at 8:43
  • @Shog9 he started the bounty before you declined, I guess he wanted an official response which you just gave by declining. Now that you also posted bounty can just be awarded to that post. (I thought you're not going to post) – Shadow The Princess Wizard Jan 13 '14 at 8:46
  • :shrug: if @Mad wants the bounty refunded, he can flag & ask for it. – Shog9 Jan 13 '14 at 8:47
  • 2
    The bounty achieved what it was supposed to do, so no need for a refund. I'm neither happy with the response nor convinced by it, but the question got the attention I wanted to attract. – Mad Scientist Jan 13 '14 at 8:53
6

This rule exists to help protect against bad migrations. If a question has no matching tags on the destination site, is it really on-topic? Is there enough overlap between the topics on both sites to suggest the folks voting to migrate it are at all qualified to judge its appropriateness? It is a fairly basic sanity-check, and given the frequent complaints about poor-quality questions being migrated, I think it is warranted. As animuson suggested in the comments, maybe y'all should work on standardizing tags?

Regarding synonyms:

Even if on the source site a tag has a synonym and that synonym exists as a tag on the target site the migration is rejected. I do not understand the logic behind this at all.

The logic is that synonyms only work in one direction. If A is a synonym of B, the question is considered to have - if B doesn't exist on the destination, then the check fails. If you want that to work, set up the synonym on the destination site - then the check for B will succeed, the question will migrate, and B will be remapped to A.

Regarding intrinsic tags:

Mad Scientist's comment reminded me that it may not be well-known that certain tags are considered "intrinsic" to a site... For instance, "math" and "mathematics" on Mathematics, or tex and latex on Tex.

An intrinsic tag never appears on any question on a site. If you try to add it, the system will prevent you from doing so, just as with a blacklisted tag... However, if you try to migrate a question with an intrinsic tag to the same site, it will be allowed (and the question will merely be tagged . This covers the common cases where a tag exists on one site while being blacklisted on another.

  • 7
    The obvious tag for a question on SO that should be on TeX or SharePoint for example would be [tex] and [sharepoint], which are likely blacklisted on the target sites because they would serve no purpose there. This check just does not make any sense to me, we generally don't take other sites into account when choosing tags, why should we suddenly have to do that? – Mad Scientist Jan 13 '14 at 8:51
  • Additionally what if the question could have a whole load of appropriate tags on the target site but because the question was wildly offtopic on the original site none of those tags were available on the original site – Richard Tingle Jan 13 '14 at 9:37
  • See my edit, @Mad. And yes, you should generally always take the destination site into account when migrating - that includes appropriate tagging as well as general-purpose cleanup. – Shog9 Jan 13 '14 at 10:50
  • 7
    I always clean up tags after the migration, on the site where the appropriate tags are actually available. I strongly doubt this check manages to catch more than the usual percentage of crappy migrations, and I don't think users should have to waste their time checking if the tag exists on the target site. The limited list of migration targets already fulfills the purpose of limiting migrations to completely unsuitable targets, I see no need for this additional controversial check. – Mad Scientist Jan 13 '14 at 11:05
  • 2
    Thanks for the answer. Just for an idea how "well" the check works and how "easy" it is to get the harmonization of tags see meta.math.stackexchange.com/questions/12307/… Compltely unjustified migration rejections due to minimal incompatibility. Sure, the tagging could be harmonized, just it feels like a bit pointless busy-work and discussion. Are there some examples where this check actually does some good? Where I look it does a lot more harm than good (namely, quite a bit and none, resp.) – quid Jan 13 '14 at 11:52
  • Here's a recent example, @quid. For what it's worth, roughly 20% of MO->MSE migrations are automatically blocked, either due to incompatible tags, due to the asker having already posted the same question on MSE, or due to the asker being suspended or quality-banned on MSE. – Shog9 Jan 13 '14 at 18:23
  • 3
    I'm not convinced that the tags are that good an indication. (Witness all the “there's a [windows] tag here so my question must be on-topic” comments.) MO/Math and CSTheory/CS are special cases because the research-level sites use researchers' names for tags while the general-public sites use common English words and phrases. Apart from that, do you have data that shows whether tags are a good indication of suitability of migrated questions? For example, for moderator-initiated migrations, how does lack of tags correlate with rejection? – Gilles 'SO- stop being evil' Jan 13 '14 at 20:04
  • For the past 90 days, the rejection rate for untagged inbound migrations on Math is 27%, @Gilles, vs. 21% for inbound migrations in total. On Tex, it's 32% vs 24%. CS doesn't get that many inbound migrations (and they're all moderator-driven), but for the record it's 20% vs 21%. – Shog9 Jan 13 '14 at 20:36
  • 1
    @Shog9 Is the volume of migration on non-SO sites really at a level where such imprecise automatic measures are necessary? I know everyone likes to complain about bad migrations, but on SE 2.0 sites the volume just doesn't seem high enough to warrant such draconion measures. The one change that truly made a difference was making rejection automatic upon closing, that saved the hassle of coordinating mods. I don't see anything else as necessary for most sites (the exception is of course SO). – Mad Scientist Jan 13 '14 at 20:44
  • @Shog9 Thank you. I don't think these figures justify a technical measure. The difference is not firmly outside the realm of statistical error. – Gilles 'SO- stop being evil' Jan 13 '14 at 20:59
  • 1
    I am not sure what you want the example to illustrate. The question might or might not be appropriate for math.SE (yet in doubt I would say actually it could be alright there). In any case that the tag-based rejection happened is by random chance. The question was very much mistagged. Had somebody cared to fix the tag using any of puzzle, algorithms, matrices (and a couple of other things) it would have worked. Conversely, there is no good reason matrix-analysis does not exist on math.SE it happens not to exist. – quid Jan 13 '14 at 21:05
  • Considering anyone with the ability to close can also retag, I find your characterization of this heuristic as "draconion" rather silly, @Mad. As for volume, define "necessary" - folks have been whining about this stuff since the volume leaving SO was tiny compared to today (and smaller than migration traffic on sites like MO->MSE, for that matter). Quite frankly, the total volume of questions where this restriction even matters is too tiny to warrant even this much discussion, much less any changes. – Shog9 Jan 13 '14 at 21:06
  • 2
    "the total volume of questions where this restriction even matters is too tiny to warrant even this much discussion" You might tell this the poor OP of a q that got closed as off-topic without good explanation as eveybody thought it would land on another site anyway, but, oh I did not because nobody thought about the tags. People do not know this. If it does not matter, than remove the check. It simplifies the software doesn't it? I would go as far as suggesting create tags rather than drop them. If people dislike them they will remove them (adding is a different thing though). – quid Jan 13 '14 at 21:09
  • 6
    @Shog9 I find it completely unreasonable to expect users to know about this restriction, and to check every time if the tags actually exist on the target site. And I'll just reverse your argument about low volume, in my opinion if the volume is low enough to be easily handled by the communtity, then any automatic measures with significant false positives are a bad idea and should be deactivated. If the volume is so tiny, it would not actually hurt to disable this flawed mechanism. – Mad Scientist Jan 13 '14 at 21:15
  • 4
    @Shog9 So you're saying that the restrictions stay there because some people tend to whine excessively about bad migrations? You're not afraid to speak your mind, so why don't you just tell those people to deal with it and base your decisions on some actual data? The problem with migrations is vastly overblown, I agree, but that is no reason for such crappy automatic mechanisms. – Mad Scientist Jan 13 '14 at 21:27

You must log in to answer this question.

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