This is incomplete
I don’t have a suggestion for the tag wiki at this moment. That will take a little more time; right now I just have some ideas about what the wiki should and should not look like. But I wanted to get this out there and get feedback on some of these ideas.
It’s not [rules]
Merely asking a question about how the rules work does not mean the rules-as-written tag automatically applies. Only a subset of rules questions would use the tag.
Thus, we reject any definition that would result in the tag being applicable to all rules questions.
It’s not [rules-lawyering]
This has been hashed out to everyone’s satisfaction previously on meta, but just to reiterate: rules-as-written and rules-lawyering are separate concepts and in most cases the tags would not even be used together, much less used to mean the same thing.
Rules lawyering is a social problem found at the game table. Rules lawyers often claim (and may even be right) that RAW is on their side, but that’s largely irrelevant to the social problem. Questions about rules lawyering typically require social solutions: how to communicate, consider others’ feelings, and so on. It usually is not actually a question about the rules themselves, since the problem with a rules lawyer usually has nothing to do with whether he is “right” or “wrong,” and everything to do with the disruption he is causing.
Thus we reject any definition that would result in the tag being applicable to questions about how to deal with rules-lawyers. Analyzing the rules in a RAW frame would only encourage a rules lawyer in thinking that lawyering is a valid and effective approach, even if he was wrong in one particular case.
It’s not [optimization]
RAW is often used in optimization, but it is not, itself, optimization or even devoted to optimization. RAW gives zero commentary on the rules, regarding optimality or anything else, really; optimization just (often, though not exclusively) uses RAW to determine how things work so that they may be then be compared for their value relative to achieving the optimization goal.
Thus, we reject any definition that would make the tag a synonym of optimization, and probably furthermore expect the definition to make this distinction clear.
It doesn’t require that the asker know the answer
This is true for all questions, and is pretty obvious when said like that, but this does have a couple of implications for RAW questions:
Some kind of RAW analysis is not required to already be in the question. While “this is my analysis of the rules; is this right?” is a valid RAW question, it’s not the only sort of valid RAW question. Questions about RAW can, in fact, be asked without any idea which rules, if any, apply.
Thus, we reject any definition that requires the asker to have already performed RAW analysis. Asking us to do so is a valid question and would be a question we would expect the tag on.
The question is not required to be one that itself requires torturous analysis of the rules from its answers. This goes back to the idea of questions where the asker has no idea what rules apply in the first place. In this case, the rules-as-written tag did not really affect the answers in any way, since we would answer the same whether the tag was there or not, by just citing the appropriate, clear rule. Nonetheless, rules-as-written is a valid tag to have on the question, since if nothing else, the asker didn’t know that when asking. Clearly a tag that requires knowing the answer is not a useful tag for someone who has a question.
Thus we reject any definition that could not be applied by a user who doesn’t know the answer to the question.
It’s not that answers are forbidden from challenging or supplementing
It is expected on our site that from time to time, an answer may “challenge” the question—in effect, say “this is the answer you asked for, but I think you are asking the wrong question, and here is why...” It is also simple reality that the rules for many games are often not good, or simply do not cover some case, and at times it seems to us answering that supplying only the RAW without caveat or commentary is itself misleading (i.e. if a rule is problematic, we might want to warn about that).
In both cases, challenging the frame or supplementing the answer to the question asked, such tangents are expected to be kept short, and they are inherently risky. We expect voters to treat such answers with initial skepticism, and that such answers have to justify their departure from strictly answering the question asked. We also do not allow such answers to entirely waive their responsibility to answer the question asked too, and generally expect that first.
Thus, we reject any definition that would make rules-as-written a unique exception to these rules. This is less a concern about what the definition is, and more a concern of how it is written. It is not necessary or desirable to invite such departures from answering the question, but we also want to avoid making the tag seem special in this regard.
It is for questions that want literal accuracy over usefulness
As I put in a comment,
As someone answering RAW questions, someone interested in what the book says and only interested in something else if the book has nothing, would be a question I am interested in answering and interested in seeing highlighted (¬_¬). From my perspective, I want to see the RAW tag if, should it be necessary, a torturous analysis of the text will be appreciated. If the asker would prefer to skip that when the rules are bad and just use someone’s recommendation, I’d rather the tag not be there. That is my rule of thumb criterion.
Thus we expect any definition we use to cover the idea that literal accuracy will be appreciated even if it is not useful for playing the game.
It is for questions that are concerned with the establishing an “objective” shared baseline
I have written a lot of homebrew for 3.5 and Pathfinder, and I have done freelance work for Dreamscarred Press, a Pathfinder third-party publisher. I am at times very concerned about the “objective” state of the rules—because as a developer whose work will be used by people I will never get the opportunity to speak with or clarify to, I need to make sure that my work fits in with the remainder of the system.
I also frequently play RPGs online, which means I play with groups I don’t know very well. RAW, even if it is stupid, gives a common baseline: it is a touchstone from which we can depart, through houserules, to get everyone on the same page with respect to the rules of the game.
Thus, for these reasons and no doubt others, people need to know what the rules are officially and divorced from even the most reasonable subjective interpretation, even if they never intend to use them. It informs their work and it forms a basis for starting conversations about the rules with those they do not know well.
Thus we expect any definition we use to cover the idea of the canonical. For instance, in the past it has been suggested that we codify what sources “count” for RAW questions; this is not actually what we want to do. More appropriate is for the definition to reference the idea of relying on the publisher’s official statements for what does and does not count. WotC did not give the D&D 3.5 FAQ the authority to change the rules, making it less than useful in RAW discussions; Paizo has given that status to its Pathfinder FAQ, making it crucial that RAW answers consider it.