Are the number of migration paths one-size-fits-all?

Server Fault is in a somewhat unique situation as far as sub-communities go (Super User may be in the same boat, but I rarely spend time there any more). Area51 has fractured our community into many smaller, but more highly specialized communities. We have: DBA, Webmasters, Unix, and Security that were all once handled on Server Fault and now have specialized communities. To a lesser extent, some content on Drupal and SharePoint might also have been on-topic for Server Fault at one point, but now has its own community as well.

I've also heard from various Server Fault mods that The Powers That Be have decreed that there will only ever be 4 migration paths. I believe there was also at least one m.so post that's authoritative on that as well.

I think that 4 is a good number for most sites. Bicycles, Gaming, Cooking, English, etc, will likely never need more than one or two migration paths. Four is probably even overkill for them! I feel that Server Fault and possibly Super User are on the opposite end of that spectrum. We have so many sub communities that have split off from us that having more than 4 migrations paths would be both beneficial to users asking the questions and users keeping the original sites tidy.

I don't think that all SE sites having the same number of paths makes sense, especially at the rate that Area51 is popping out new sites with overlapping content.

TL;DR - I don't think that the number of migration paths are one-size-fits-all

Arguments against adding additional paths

When I've brought this up in the past I've heard a few arguments against adding targets that I'd like to preemptively address:

Only a small percentage of questions are migrated to sites outside of the migration path. It's not really needed.

It's not fair to assume that users will flag and type out an explanation of why something should be migrated somewhere in the path (and risk losing fw) when they can just close it outright as off-topic. There are a lot more users on SO, so I'm sure this isn't a problem for you, someone will likely flag something for migration to the right site. On SF, that's just not the case.

If you increase the number of options, the percentage of questions migrated to those sites will increase.

People are stupid! Too many migration paths will confuse new users!

The threshold for voting to close is 3k. Users should have a good grasp of things by then. I do agree that this could potentially open the door for an increase in "bad" migrations, so I propose a tiered voting system: At 3k, you can vote for the 4 standard paths at 10k you can vote to migrate to a larger set of sites.

This should allow most questions that belong elsewhere to get there safely without much downside. By the time a user hits 10k, he or she should have a good grasp on how we work.

Where do we draw the line for what sites should and should not be in the migration path? It seems like a slippery slope!

Well, it is a slippery slope, but it's one that we're already on. With the success of Area51 and all of the sites that it's launched, we're facing this problem already! Server Fault gets a lot of Unix questions, because people assume we're the best place to ask (and we were at one point), but if they're not System Administration related, we don't want them any more. Times change, we need to keep up!

Area51 is a great thing. We need to make sure that use old foggies from the original trilogy sites can keep up with the times. Adding relevant migration targets helps promote these new communities as well as keeping the original site on-topic.

I think that if a general consensus on each site's meta is that a site should appear in the migration path for 10k users, then we should make it so.


Pls let us haz moar sites in da migrashun list. Even if it's only for 10k users.

I know that The Powers That Be have been adamant in the past about the number of migration targets in the list, but I think that it really needs to be reconsidered in light of the success of Area51 and the new communities that overlap with us old-timers from the original trilogy.

  • 3
    It's not that users are stupid, it's that people in general don't like choice. More choices == hesitation. It's a psychological effect that we have to keep in perspective as sysadmins (and probably in other fields.) If you keep things simpler, you will have more participation and less hesitation from people...also people tend to go with the fail-safe or default options presented to them, assuming that you or the system "knows better" if they lack domain knowledge of the problem. – Bart Silverstrim Jan 11 '12 at 16:18
  • In other words, if a question doesn't belong here, participants may know it doesn't belong here, but they won't know where it actually belongs, so they'll just leave it to the system to sort it out or get discouraged from participating. – Bart Silverstrim Jan 11 '12 at 16:19
  • @BartSilverstrim I agree that it may be the case if a user is overwhelmed, but by the time a user hits 3k they should have a decent understanding of Stack Exchange as a whole, hopefully. Also, I suggested that the additional migration paths only be available for 10k users to further combat this hesitation. I'd really hope that 10k users wouldn't be overwhelmed by having 2 or 3 extra choices. – MDMarra Jan 11 '12 at 16:21
  • 2
    ♥ the tiering idea! – MikeyB Jan 11 '12 at 16:21
  • I hadn't addressed the tiering idea, just the connotation that the users are idiots by default by pausing with the options :-) – Bart Silverstrim Jan 11 '12 at 16:22
  • 4
    @MikeyB I stole it from ChrisS, which is where I get about 90% of my good ideas. – MDMarra Jan 11 '12 at 16:23
  • You stole some of this from discussing in chat! ATTRIBUTION LICENSE, MAN! – Bart Silverstrim Jan 11 '12 at 16:24
  • 2
    I for one would love to see more migration paths, SO could use this as well (and don't buy the "people don't like choice" argument...people don't like having choices constrained where the correct answer is unlisted) – kekekela Jan 11 '12 at 16:50
  • 2
    @kekekela: It's not something made up. amazon.com/Paradox-Choice-Why-More-Less/dp/0060005688 or Google "paradox of choice" – Bart Silverstrim Jan 11 '12 at 16:59
  • People don't like choices for things they're actually interested in. If it's not a core interest, they don't want to think about it. – Bart Silverstrim Jan 11 '12 at 17:00
  • 1
    @BartSilverstrim I agree people only want to see "core interests", but then the solution is "List all sites, ordered by frequency of migrations, with a see more link after the first N to show the rest", not "Display N sites. If it's not there you have to do extra work and manually flag it - tough rocks." – voretaq7 Jan 11 '12 at 17:28
  • 1
    If you order it by frequency of migrations, you won't have a consistent list, which could further confuse people used to finding things in approximately the same area...just something to consider. – Bart Silverstrim Jan 11 '12 at 17:30
  • 1
    Luckily a new user can't vote to migrate so we don't have to worry about that. – kekekela Jan 11 '12 at 17:48
  • 1
    A single dropdown could handle it, you act like its got to be some huge space clogging behemoth that is literally going to terrify grown men. Claiming "paradox of choice" here is like saying we shouldn't have dropdowns for State on address forms. – kekekela Jan 11 '12 at 17:50
  • 2
    Even on SO most of my flags are "please migrate to xyz.SE" because the list is insufficient. – CodesInChaos Jan 11 '12 at 20:59

First, some stats:

In the last 90 days (data gathered on January 15th 2012), Server Fault has had 597 question migrated away from it:

  • Super User†: 332, 12% rejected
  • Stack Overflow†: 74, 10% rejected
  • Unix & Linux: 47, 14% rejected
  • DBA: 35, 8% rejected
  • Apple: 32, 0% rejected
  • Web Masters†: 27, 7% rejected
  • Ubuntu: 20, 5% rejected
  • Security: 18, 0% rejected
  • SharePoint: 7, -- not enough data
  • Programmers: 2, -- not enough data
  • WordPress, Web Apps, Project Management, Electronics: 1 each, -- not enough data

†Existing migration targets at time of writing.

Tossing out the rejections, this would mean that approximately 154 legitimate migrations were not covered by our migration targets (compared to 384 that were).

Note, there's about a 10% chance the migration is rejected (closed on the destination site); this is important.

To get a feel for how many questions are slipping through the cracks due to missing migration targets, here's a Google doc of the last 50 off-topic questions that were not migrated and have not yet been deleted‡ (+ if they should have been migrated).

Of those questions that could have been (I hesitate to say should, as they were almost universally a bad and not migrating garbage is 100% correct even if the question is nominally on topic elsewhere) migrated, overwhelming they would have been moved to Super User, after that Stack Overflow, and then a smattering of edge cases (Unix/Linux & Ask Ubuntu in my opinion).

‡So everyone can see them; this excluded 2 questions from the sample.

Conclusions I'm comfortable drawing from this investigation:

  • we're not dropping many migrations
    • essentially everything that should be migrated ends up migrated, it's a question of voting or flagging not end results
    • corollary, we're still migrating some things that should be (~10% rejection rate)
  • Web Masters probably shouldn't be a top level migration target
    • obvious candidate to replace it is DBA
  • Unix and Linux might need to be a migration target
    • there's an open slot, it gets about 2/3rds the legit migrations Stack Overflow does

Assuming Unix & Linux was already a migration target, we'd have seen approximately 1.3 "resulted in migration" flags per-day on Server Fault (that's including those ultimately rejected by the destination site) over the last 90 days. This seems reasonable, given how murky the on/off-topic status of a question can often be.

I'm very much against gating access to migration targets based on reputation, it doesn't follow in my opinion.

Consider, at 3K Server Fault reputation you're expected to know what is on-topic for Server Fault. You can vote to close with a curated list of "good chance" targets for the migration case.

At 10K reputation... you've got 7K more experience with Server Fault but nothing with any of the other sites on the network. What's changed that would justify more migration targets? It just doesn't follow that higher rep on Site A implies understanding of Site B (Jon Skeet would be omniscient by now were that the case).

Moderators have the recourse of discussing flagged questions with moderators on other sites, which side steps the whole per-site concerns. Of course, moderators being human, mistakes still happen.

I'm also just generally against increasing the complication of the migration system; increased complexity drives existing users away (at least from the feature), and makes Stack Exchange even harder for new users to approach than it already is. If we had strong evidence the current system wasn't working we'd have no choice to apply tweaks (or overhaul the whole thing), but what I'm seeing doesn't make me feel anything drastic is necessary.

Based on this data, the Community Team reached out to the Server Fault moderators to see if they were OK with changing the migration targets.

The majority were behind it, so we've made the migration target changes.

enter image description here

  • 2
    I'm glad that we have some real hard statistics on this. I think that the last 50 OT w/ no migration might be a bit small of a number to be statistically relevant, but your points are well taken. I just hope that as Area51 continues to be successful and fracture the larger communities, the SE team will keep an open mind to lengthening the list as-needed. – MDMarra Jan 23 '12 at 19:47
  • 1
    I'm generally pleased with this change -- I think it is at least an 80% solution. We should probably revisit the migration statistics in say 3 months and see how this pans out... – voretaq7 Jan 23 '12 at 19:48
  • 2
    It's possible that these data may be flawed because they were gathered while the original system was in operation. If an average SF user thinks "this belongs on Ubuntu," then sees that Ubuntu isn't in the migration list, several things could happen: he might submit a mod flag asking for migration; he might not know about mods' "migrate anywhere" power and submit an off-topic close vote, or vote for Unix and Linux instead, or just give up; and he might be too lazy to exit the close menu, enter the flag menu and type out a custom mod flag. I'm interested in what effect the change will have. – Pops Jan 23 '12 at 19:55
  • 1
    @PopularDemand The coexistence of U&L and Ask Ubuntu really baffles me. I can understand (even if I disagree with) other sub-communities, but those two existing independent of each other seems just plain weird. – MDMarra Jan 23 '12 at 19:57
  • 1
    What are the conditions by which something is labeled as a "declined migration"? I would assume that means closed, but would closed as Exact Dupe count as a decline? – Scott Pack Jan 23 '12 at 19:57
  • @MDMarra That was addressed in this blog post. – Pops Jan 23 '12 at 20:05
  • @PopularDemand While I understand the sentiment in that blog post, I'm with MDMarra -- From a strictly technical standpoint it's hard to justify that level of fragmentation. By that logic U&L shouldn't even exist -- It should be AskFreeBSD, AskNetBSD, AskOpenBSD, AskRedHat, AskAIX, AskHP-UX, AskSolaris... – voretaq7 Jan 23 '12 at 20:13
  • @voretaq7 You're not preaching to the choir, but you are preaching to... someone who doesn't listen to choral music? I don't use AU or Unix SE, and I didn't take part in the fork debate. I merely remembered that it happened, so I provided a link. – Pops Jan 23 '12 at 20:17
  • 1
    I'm not preaching, just grumbling about faux-distinctions that make no technical sense. Like Win98 versus WinME :) – voretaq7 Jan 23 '12 at 20:23

This is particularly a problem for ServerFault: We receive an avalanche of questions that are absolutely not about system administration, including:

  • Programming questions (-> StackOverflow, check)
  • "How do I get my desktop to do X" questions (-> SuperUser, check)
  • "How do I do X on my Linux desktop?" (-> Unix & Linux, missing)
  • Questions specifically about databases/DBA (-> DBA, missing)
  • Questions about website stuff (-> Webmasters, check)

The net result of a lack of migration choices is questions being closed as Off-Topic: Off Topic ("Go away, we don't want your kind here!") rather than being migrated to a site where the user can get help.
(If we want to talk about the psychology of choice let's also talk about the path of least resistance: It's easier to vote to close as off topic than to flag and type in "Probably belongs on Other Site")

I think the current method of picking the migration targets is probably fine, but the proliferation of StackExchange sites seems to require at least one additional migration "slot", if not two.

  • 6
    Don't forget about Security! Most of those are on-topic in both places, but the fact remains that there is also an overlap there that we aren't able to handle without flags.. – MDMarra Jan 11 '12 at 16:38
  • 1
    There is one slot open on Server Fault. You should get the mods to check which is the most common target not on the current list and get it added. It's not a complete solution but would help. – ChrisF Jan 11 '12 at 16:41
  • 3
    @ChrisF This was already brought up on our meta, which raised this larger question. Instead of only partially solving the problem by cramming the square peg into the round hole, I'd like to redesign the hole altogether. – MDMarra Jan 11 '12 at 16:43
  • 1
    @MDMarra - Oh I agree. However, it might be that a partial solution is better than no solution at all. – ChrisF Jan 11 '12 at 16:44
  • 1
    I could absolutely go for a partial solution to fix the Now Problem, but I much prefer general-case solutions that make the problem go away forever :-) – voretaq7 Jan 11 '12 at 17:40
  • @Chrisf The meta-post in question: meta.serverfault.com/questions/2582/… – sysadmin1138 Jan 11 '12 at 18:21

I can't help thinking 5 migration targets like we have now is an arbitrary choice as it is - why not 4, or 6?

In any case, I agree that the current "5 and damn the torpedoes" option is not great. The Stack Exchange team can't have it both ways; If the sort of fragmentation and splintering that Area51 can cause is seen as a good thing, then tools that scale to cope with the side effects of that fragmentation must also be a good thing.

I think there are certain "families" of sites within Stack Exchange, for example the computer related ones, and I'd like to see us able to migrate between all those sites when they are out of beta. A "tree" migration system might help prevent the menu from getting too large, though at the risk of hiding some options from people who aren't prepared to look for them (but then I suspect these people are already not flagging and asking for moves to unlisted sites).

  • 2
    Good "Comment", bad "Answer" – Chris S Jan 11 '12 at 18:08
  • 1
    Good "Answer".. – kekekela Jan 11 '12 at 18:43

Say a 10k user votes to close an SF question and uses this new tiered ability to recommend migration to IT Security. What happens if the next four visitors to that question are 5k users? Will they now be able to vote to migrate to IT Security as well?

What if two 4kers show up before a 13ker and think "I wish I could migrate this to IT Security, but I'll have to settle for 'Off Topic' and a flag"?

I'm not necessarily against this proposal, but I'm not sure how to work out its side effects.

  • I had the same thoughts you did -- I came up with a "solution" though I'm not sure how much pain and suffering it would inflict on the developers... – voretaq7 Jan 11 '12 at 16:44
  • 1
    ...is that a bad thing, @voretaq7? – Bart Silverstrim Jan 11 '12 at 16:57
  • 2
    Why not just display the expanded list for all 3K + users. I'm sure they'll manage the slightly expanded choice. – Adam Rackis Jan 11 '12 at 17:18
  • 1
    @AdamRackis that's been rejected before (though that hasn't stopped me from suggesting things in the past). – Pops Jan 11 '12 at 17:42
  • 1
    @AdamRackis Ultimately I think that showing everything is the best choice but I don't know that 3K is a high enough threshold for that kind of power. We definitely don't want to open a floodgate of bad migrations because eager 3K users start flinging stuff around based on site title, and in my experience higher-rep users seem to be more engaged (peruse the site's FAQ to make sure it's a good fit, hop on a site's chat and ask if the fit is questionable) – voretaq7 Jan 11 '12 at 17:44
  • @AdamRackis That was what I originally was throwing around in SF chat, but we hate "bad" migrations on SF as much as anyone else. I think my making the non-standard choices a little harder to get at, it will keep them to a minimum. – MDMarra Jan 11 '12 at 18:04
  • @PopularDemand I don't think that a 10k user's vote should open the way for 3k users to choose that site as well. That would lead to an inconsistent voting experience. For a question to be migrated to a site, it only needs a majority of the votes to be in favor of the migration, not all of them. Since 10k+ users tend to be more active, I think that this issue will really take care of itself. Perhaps all Off Topic votes without a migration path chosen can be considered votes for the highest voted non-standard migration path? Any other ideas on how to keep the experience consistent but usable? – MDMarra Jan 11 '12 at 18:12
  • Migration requires four out of five close votes to agree on a destination site, not a simple majority, @MDMarra. More importantly, I don't have faith in your "10k+ users do most of the close voting" assumption, but it'd be interesting to see numbers from the team on that. – Pops Jan 11 '12 at 18:23
  • @PopularDemand Hm, I thought it was 3/5 for the migration. That's good to know! I think that the answer becomes if 2 10k users vote to migrate to the same non-standard site, then all off-topic votes with no migration target from <10k users should count as votes toward that non-standard site. This, of course, gives 10k users a "stronger" migration vote, but high-rep users typically get more votes/responsibility anyway. – MDMarra Jan 11 '12 at 18:26
  • It used to be 3/5, but the team changed it after seeing too many bad migrations, @MDMarra. – Pops Jan 11 '12 at 18:59
  • If the tiered system is too messy, there's always the easy option of just listing the sites in the default close dialogue. – MDMarra Jan 11 '12 at 19:07
  • 3
    @PopularDemand It's 4/5 for SO, still 3/5 for everyone else. See meta.stackexchange.com/a/97700 – Shane Madden Jan 12 '12 at 2:24
  • Right you are, @ShaneMadden. – Pops Jan 12 '12 at 14:02

Alternate idea instead of adding more migration radio buttons -- Keep the migration list as-is for users below a given threshold (10K?).

For users above the threshold present the current migration list, plus a list of all the live StackExchange sites for them to select from. If one of those sites is selected as a migration candidate replace the lowest-ranked site in the list for users below the threshold.

This sounds like a lot more programming work than adding another slot or two for migrations, but addresses the choice paralysis issue Bart brought up and lets everyone with vote-to-close privileges participate in the migration decision while also giving more experienced users a broader range of migration choices.

  • 3
    One problem: I'm a 10k user on SO, earning almost all my rep on C++ programming. How would I know what is on topic for dba or bicycles, or what is the difference between Wordpress and drupal? Never been there... – Bo Persson Jan 11 '12 at 17:13
  • @Bopersson - that's a whole other debate. There are lots of people on SF and SU who would suggest that not knowing what's on topic on those sites doesn't stop SO regs from migrating in those directions... And I'm sure some equally dodgy migrations have gone the other way too. – Rob Moir Jan 11 '12 at 17:27
  • 1
    @BoPersson This obviously assumes that users at or above $threshold are smart enough to not randomly throw around migration votes: If they're unsure we need to trust them to either leave the question alone or flag it for mod. I'm not normally one for trusting users, but I don't see any good alternative in this case. (You can also make the same argument with the current migration tools -- the typical solution is to hop on the other site's chat and ask if they want the question) – voretaq7 Jan 11 '12 at 17:35
  • 1
    I want to be clear that 10K is also just a number I threw out there - it's a nice round landmark that active and responsible users can get to without too much trouble (at least on SF). It may make sense to have this a 12k or 15k permission (or higher, or lower, it's not easy to pin down) – voretaq7 Jan 11 '12 at 17:39
  • My similar suggestion was not well-received (though I didn't have the rep criteria): meta.stackexchange.com/questions/117581/… – Jacob G Jan 11 '12 at 17:58
  • 1
    @BoPersson That's a user problem, particularly on SO. If you aren't sure where a Question should go, do not vote to migrate. By the time a user hits 10K they should know that (though it seems there are plenty of counterexamples). – Chris S Jan 11 '12 at 18:04
  • 1
    @voretaq7 - I'm actually at 18k so the limit could still be a problem. :-) And I do indeed most often vote to close as "Off topic" without selecting a specific site. Having more choices wouldn't help me there. – Bo Persson Jan 11 '12 at 18:12
  • @BoPersson I think there will still be a lot of Off-Topic closes (if only because of that uncertainty about if migrating a question is appropriate), but I know I've looked at questions and said "This belongs on other site - I wish it were easier to migrate it" – voretaq7 Jan 11 '12 at 19:20

...what about adding migration lists based on the tags the question has? Include some common defaults but tack on one or two "probable" migrations based on that...

I liked the tiering idea, but I wondered if it's not something that would be imbalanced (i.e., if the question belongs in XYZ instead of here, the user should have some way of getting it in the apropos spot, even if they're not an overly experienced user.)

I can understand it being kind of like another unlocked achievement, but here you're talking about a fundamental tool for keeping sites clean and proper (and getting the "right" eyes on the problem for a solution) so maybe using a tiered gaming system isn't apropriate (or it may allow a question to be migrated to yet another wrong site before it finds a proper home.)

  • This has been brought up before. I can't find it right now though – ChrisF Jan 11 '12 at 17:25
  • 1
    Tags are ambiguous outside of context. Java on SO means something entirely different than it does on Cooking, for example. That's probably an extreme example, but I'm sure there are hundreds more when you mix technical with non-technical communities. Just think of all the overlap that everything will have with unrelated sites like English. – MDMarra Jan 11 '12 at 19:12
  • Perhaps, but ideally there would be little overlap within context. Why would someone insert a question about coffee (cooking) in Serverfault? It would be obvious that the tag would be technology related, so it should go into a technology-related migration. – Bart Silverstrim Jan 11 '12 at 19:33
  • It's obvious to us (humans) given context and heuristic capabilities, but an algorithm (at least a dumb one: "Show all the sites that have a java tag") wouldn't know that -- It would offer up Cooking honestly "believing" it to be a good choice. – voretaq7 Jan 12 '12 at 5:48
  • @voretaq7: I was thinking that the heuristic wouldn't just be the tag, but the context of the tag. If the question was posted in SF first with a Java tag, surely the algorithm can say that if SF is categorized as "technology" then Java under SF would be programming, not cooking. The inference would be based on categories of the sites as well as the tag to form a context so Jave in SF could mean SO or SU or Programmers, while Java in Cooking could mean migrate to Travel (is java in Java safe to drink?) – Bart Silverstrim Jan 12 '12 at 13:29
  • @BartSilverstrim A less naïve algorithm certainly improves things. This would require some kind of metadata about the sites though (I'm not sure if the stack network stores that, or if they'd want to) – voretaq7 Jan 12 '12 at 16:08

Another suggestion to add to the growing list:

How about giving users with the "Cast close and reopen votes" privilege to vote to migrate to any other site where they have that same privilege?

For example, if I have enough to rep to cast migration votes on Stack Overflow, AND I have enough rep to cast votes on e.g. Cooking (for example) then it's implied by my privileges that I know enough about BOTH communities to understand whether a given question on one of these sites is on-topic on the other.

In fact, come to think of it, wouldn't it be better if the existing radio button list did this rather than use a hard-coded list of sites? I've read complaints in the past about people migrating bad-subjective (as opposed to good-subjective) questions to Programmers, thinking that they would be on topic there. This suggestion would presumably cut down on occurences of this problem.

  • I kind of like the rep threshold idea in general (it would probably cut down on bad migrations), but as much as I peruse other sites (like DBA and IT Security) I'm really only active on one. Building up 3k rep is a hurdle I'd have to jump, but I guess it's not a bad one. – voretaq7 Jan 11 '12 at 17:48
  • 1
    How about leaving the default choices there, but make an option to select any site a 15-30k privilege. I am not sure a new 3k user is really ready for that ability. – Zoredache Jan 11 '12 at 18:11
  • @voretaq7 - A great point. Lurking on a site can give you an understanding of it's remit, but not generate you any rep. – razlebe Jan 11 '12 at 18:16
  • @Zoredache - Having 15k rep on SO doesn't necessarily mean you know what you're talking about on Cooking, though. That's why I've proposed linking the ability to the voting privileges, as by definition you are considered reputable enough to migrate-by-voting when you have that privilege. – razlebe Jan 11 '12 at 18:17
  • There's also the question of collecting enough additional votes (You vote to move something from SO to Cooking, but do we now show Cooking as an option on that question for everyone without the off-site rep?) – voretaq7 Jan 11 '12 at 19:22
  • @voreteq7 - I guess I envisage everyone else seeing an "Off-topic" vote if they don't have the rep on cooking. Perhaps they could see the cooking vote, but not add their own unless they have the cooking rep. I take your point though - the system is complex. – razlebe Jan 11 '12 at 19:24

You must log in to answer this question.

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