[moberemk]

I generally agree with all your points, but I also remember the forum posts and I find that SO has at least one big advantage over those: it's really quick to pick out a good answer compared to a long thread. Just the fact that a) answers are clearly demarcated vs comments and b) the author can select a "correct" answer makes them way more skimmable than, say, a phpBB thread where every post looks the same.

[SCdF]

I definitely see that side of those, though I think they both have pros and cons.

Especially in the JS space[1], there are lots of historic answers upvoted on SO that are now wrong and you have to scroll around (or find a slightly rephrasing of the same question that occurred more recently) to find a good answer. Unfortunately the longer a question is around both the more answers it has and the more likely the answers are to be wrong. On forums / mailing lists the question would just be reasked. On SO it is supposedly a wiki so new versions of the same question gets closed.

[1] Because it has evolved a lot, Rust has similar issues, as I'm sure lots of languages do

[shepmaster]

> Rust has similar issues

I really do try my hardest to keep old questions and answers relevant. Stack Overflow allows editing Q&A for a reason, and any time I stumble on an older post, I at least think about updating it.

One thing that I've occasionally found frustrating is that someone sees a question was asked 2 years ago and thus assumes that it must be invalid / out-of-date. They simply don't look at the edit dates.

Remember that Rust has a pretty strong backwards compatibility guarantee. Any answer using only the standard library in the last 3 years should still be valid (crates are a different concern). If you find a question you think has become stale, add a bounty to it to raise attention — that's the SO recommended path of action.

We also have a Rust chatroom on SO where a bunch of regulars hang out and people are welcome to pop in and ask if a Q&A is still valid.

[lovich]

Should the answers be edited? Everyones had to work with software on the job that wasn't the latest edition. If we edit answers to software after every update doesn't that just fuck over everyone who's not capable of updating?

[shepmaster]

As your other responses have stated, editing doesn't necessarily mean "destroy the old and replace it".

A lot of the editing I do is to improve the grammar, reduce fluff, use Rust-standard indentation, improve the formatting, include complete error messages, update links, etc. None of that should affect users of older versions other than to make it easier to get to the core content of the Q/A.

When a new Rust feature comes out, usually the original part of the answer gets a header denoting the compatibility. https://stackoverflow.com/a/28953618/155423 is such an example.

Note that SO does keep a revision log of edits (e.g. https://stackoverflow.com/posts/28953618/revisions) to a post. You can browse that if you think there might be something hidden that's worth exploring. This doesn't necessarily help with search engines, of course.

[dasil003]

I often update my answers over time to cover multiple versions, usually as the result of someone leaving a comment pointing out a problem. It's pretty easy to do in a way that preserves the old information (eg. https://stackoverflow.com/a/1576221/8376).

[acjohnson55]

You shouldn't necessarily erase history, but it's pretty easy to move your old text under an Old Answer banner or something, if you still think the info is relevant to someone or even that surfacing the delta between versions is educational.

[SCdF]

To be clear, Rust is easily 10x better than JS in that regard, it was just an obvious example for me to use where <1.0 code is now weird and wrong :-)

I have nearly given up looking for JS answers in SO because they are such garbage these days. I am learning Rust right now and still find answers on there useful.

[shepmaster]

> where <1.0 code is now weird and wrong

We actually have a specific tag for those cases (`rust-obsolete`), so if the question / answer cannot be rehabilitated for Rust 1.0, that's a useful thing (I also try to add an in-question disclaimer to make it easier to spot)

[gdulli]

> there are lots of historic answers upvoted on SO that are now wrong and you have to scroll around (or find a slightly rephrasing of the same question that occurred more recently)

And how many of those slight rephrasings never got a chance to get a more relevant, correct, modern answer because they were marked as a duplicate?

[jodrellblank]

Under the StackOverflow-as-Wikipedia model where it's a card catalogue of best answers, the right thing to happen is:

- they get marked as duplicates, so that all future people coming in via Google get funnelled to one place

- that place gets new modern, more relevant answers

- they get voted up because they're useful

- people add comments about answer compatibility

- they either overtake the other answers and the other answers stay around as history, or if they don't there is a single place where relevant answers can be found.

1 question with 10 answers is a lot more discoverable than 10 questions with 10 answers.

One leads to 9 answers to skim-read and ignore and 1 answer to use.

The other leads to you finding 6 questions and 6 answers you don't like and 0 answers to use and missing 4 you didn't know existed.

[walshemj]

Not in my experience on the workplace so - in some cases the answer the OP needs is "NO do not do that", but the highest rated answer is long one that really does give the answer that is needed.

Sometimes you have to work out the underlying Q is and answer that even if the OP doesn't want to hear it.

[boomlinde]

I much prefer the answers that explain why, if you really shouldn't do something in addition to an answer to the actual question. Usually there are perfectly valid reasons to do things you "shouldn't" and until you know that there isn't one you should just answer the question.

[Can_Not]

> Sometimes you have to work out the underlying Q is and answer that even if the OP doesn't want to hear it.

And often times you have people trying to "work out the underlying Q" leaving your question still open but with useless answers.

[marcosdumay]

> a) answers are clearly demarcated vs comments

About that, there seems to be a recent trend that answer quality is so low that the only actual useful information is on the comments.

[tneu89n43nth]

I have the exact opposite experience. In my experience the quality of answers has been improving significantly recently.

[daphneokeefe]

I believe that you can't be down-voted, and therefore lose your hard-won points, by commenting. But if you answer, you can be down-voted. So for some participants, commenting is prudent.

[mark-r]

And then there are the poor souls who have a useful comment to make, but can't because you need reputation points to make comments. So they leave an answer instead, which is completely inappropriate and just leads to more discouragement.