While there is at least one potential technical advantage to having subsites (using DNS for load distribution) and well-defined sub-communities have significant social advantages, subsites have significant question targeting issues.
Using tags, there would be fewer off-topic questions and “migration” would be a matter of retagging. Tags could be used to provide filtering by requested answer accessibility (newbie questions irritate some and excite others, expert questions are unintelligible to some and precious to others).
Tags might also be used for speculative degree filtering. (At Stack Exchange, some subsites tolerate or even encourage reasoned speculation [usually necessary for something like World Building but sometimes useful even for more mature technical topics] but some aggressively close speculative questions even when phrased in an answerable manner recognizing the need for speculation.) Other mechanisms for such filtering might be more appropriate.
Tag-related reputation could serve a similar role to subsite reputation and there would not be the issue of “association bonus” for those who have proven that they have basic good behavior.
It might even be possible/practical to have tag-set themes similar to sub-site themes with users providing themes and potentially even contracting artistic experts to design excellent themes for a favorite tag-set. Site “graduation” has been an issue at Stack Exchange; allowing a less popular topic to be supported as much as the users wanted might be better than centrally controlled specialization.
I do not know if such would be practical or even if it would be beneficial. This loose proposal might provide insight into a practical alternative or the subsite design might be the best option despite its weaknesses.
One of my areas of interest and (some) knowledge is computer architecture, which has some electrical engineering aspects, some software development aspects, some general less-mathematics-oriented computer science aspects. This mildly cross-disciplinary topic is somewhat fragmented on Stack Exchange (SuperUser has some questions, Electronics has some, Computer Science has some, StackOverflow has some [mostly technically off-topic]) and questions often enough get closed as off-topic with no migration. Other topic areas are probably even more cross disciplinary.
Tags are already going to be used, and each site will have their own, and they will use the same tag names with different meanings.
Furthermore, I think it is important that each site have clearly identifiable borders, because each site will have its own community and its own community defined rules.
Different sites also have their own sets of tags (tag clouds) – some of these tags describe site-specific topics, and some entail specific handling (behaviour or rules) defined the community.
There’s also the issue that the point of having a community (or several communities) is so that questions get answered. According to Joel that distinguishes SE from other sites (e.g. Yahoo Answers or whatever, where it’s easy to ask but there is nobody to answer).
A community might define a site around a specific topic or range of topics, and be willing to answer questions about that – conversely if you don’t know which site (which community) you’re questioning then the question is less like to be answered.
On the one hand, fewer off-topic questions is good.
On the other hand, we’d need to either implement custom close reasons per major tag, or have a lot of custom reasons that would mostly be irrelevant, or do a lot of compromising on wording to cut the set down to a manageable dozen or so.
Yeah, I don’t think meta tags are a good choice for this kind of separation. For that matter, sites are probably not a good arrangement either, at least not for a newbie-expert divide. If you slice up your answering community with significant barriers between different levels of expertise, you lose the progression of skill that can otherwise allow someone to steadily expand their range of answerable questions as they learn more. Worse, because it’s difficult or impossible for a newbie to reliably determine whether their question is “expert” or “newb”, you substantially reduce answer quality: experts can’t easily answer “newbie” questions that actually turn out to need them, and moderately-skilled answerers can’t easily answer misjudged “expert” questions. A good example of this is the English Language Learners/English Language & Usage “split”. This is often misunderstood as a real-world example of a newbie/expert divide working well. But in fact it actually demonstrates the only real way you can make such a split work: divide based on audience interests and goals, not level of skill. There are experts in second-language acquisition on ELL that can explain to even someone quite fluent in English how some particular point of the language makes sense, when a native speaker won’t need such an explanation past the age of 5: it’s picked up without thinking about it. Contrariwise, there are some fairly simple questions on ELU that aren’t especially germane to an ESL learner or teacher, such as basic explanations of Early Modern English grammar.
As I understand it, something very like tag reputation is desired, though it may not be a simple accumulating rep. It’s not in the spec yet, though, and probably isn’t MVP.
Sorry, but no. One serious problem with this is that communities around different disciplines have different norms for talking to each other. Having distinct sites allows the rules to be customized to each. For example, artists might want to talk about the “feel” of something, which would be way off topic on an engineering site. On the other hand, telling someone in an engineering context to read the manual, or even using the common short form “RTFM”, isn’t really unusual, whereas it might be taken as a serious insult in an artistic context. Trying to decide what rules apply from tags doesn’t work because it’s not about the content of the post, but the intended audience.
Another problem is that common words can have very different meanings depending on the community. Take “shield” for example. It means one thing to historians, something completely different to electrical engineers, and something completely different again to users of the Arduino microcontroller development boards. However, each community might want a “shield” tag.
Yes I think (a site’s or community’s) focus around some topic is beneficial. Your argument about a “computer architecture” question being aimed at individuals within several other communities becomes less plausible when you also consider other topics (i.e. sites, communities) like Politics and Parenting and Pets, and the religion sites, etc… they rather benefit from being separate[d] IMO.
That said, a[ny] community might presumably be welcome to define “their” topic as broadly as they like.
Perhaps even SE isn’t sure whether the other software sites should be separate from SO – and those sites weren’t initially but evolved to be so.
Plus I think you’re right, that the use of tags could (and possibly should) be generally more powerful/capable somehow – in theory that might be a killer app eventually. I haven’t seen many examples (of tag implementations other than SE’s) though.
Tags … and/or user distribution lists, like people used to subscribe to list-servs.
And possibly you want to reconsider how topics/sites are initiated – have like an Area51 for tags. So if the OP wants a new tag computer-architecture then they might do so, and begin to use it, and existing users of other sites might subscribe to those tags – so that they then begin to computer-architecture questions and meta-discussion.
Defining topic-hierarchies though is a hard problem, possibly AI-hard.
Alternatively, would it be possible to migrate questions from one community to another? That could reduce a lot of people closing questions as “off-topic”, when the user might just be overwhelmed with the number of communities.
Migrating is pretty tricky in the sense that the tolerance for mistakes are much lower. You don’t want a war between different sites. If a question should be migrated, it should be 100% sure that it is on topic on the site it gets migrated to. Furthermore, it should be good or at least descent question according to the other sites standards.
The above are my prejudices about what other sites think when they get questions dumped from SO. I have nothing of substance to back it up. Correct me if I’m wrong.
I imagine that a better solution is to just inform the user that site X might be a better choice and letting the user to repost themself.
Agreed. I’ve thought for a long time now that migrations on SE were a bad idea. It’s up to the user to pick the right site. That’s part of the job of doing the minimum research and writing an on-topic question.
Not having migrations simplifies things, and avoids the drama around bad migrations on SE.
Those writing answers also have some responsibility to not answer something that doesn’t belong on the site. Answering something that ultimately gets closed is bad for the site, so some dis-incentives to do that are helpful. Having your effort answering the question get lost is one of them.
Meta on Codidact is just another category on the same site, so “migration” is much lighter weight. (We haven’t specified a trust level for this – missed it.)
I think the trust level needed should be configurable per community, and possibly even per category (for example, moving from a sandbox category to the main category should need much less trust than moving from main to a canonical questions category).