12 ms·
One challenge not discussed here is when an accepted and highly voted answer _used_ to be a good answer, but no longer is in later versions of whatever software
by jrochkind1 8y ago
One challenge not discussed here is when an accepted and highly voted answer _used_ to be a good answer, but no longer is in later versions of whatever software/platform/library is discussed. This is also something that gets worse over time, like the uniqueness problem.
I think it's worse than the uniqueness problem, multiple duplicate _good_ questions/answers don't actually harm, misguide, or delay anyone looking for an answer; one _wrong_ one does.
And of course, if you try to ask a question again because you think the old one might be outdated (if you were _sure_, you wouldn't have to ask probably), it'll likely get closed as a duplicate. :)
- manigandham 8y agoRemove the entire concept of "accepted" answers and let votes decay or be withdrawn, along with scoring the rankings using time as an additional input. Quora has recently introduced secondary voting mechanisms where answers will randomly have a "is it still relevant" toggle for users to continue providing signal beyond the initial vote.
- rc_kas 8y agoI think its harder than that. Some anwers are just timeless. Others should decay.
- tobr 8y agoIf they’re timeless they should continue to get votes, no? Even if everyone stops voting in a thread all the votes would decay at the same rate, so the timeless answer would stay at top.
- Piskvorrr 8y agoI can confirm this: I'm getting votes from stuff I have written in the Permian, Triasic and Jurassic periods, respectively.
- manigandham 8y agoDecay will apply to all answers so it'll still stay at the top of the rankings unless a better answer comes along. Tons of options but better than the static scores we have now.
- bryanrasmussen 8y agoIf a better answer comes along in a question with 20 answers already it probably won't get up to the top anytime soon. If a better answer comes along in a question that is then marked as a repeat of a question that has a worse answer, forget it. At any rate if a better answer comes along it should rise to the top in the matter of <24hrs.
- manigandham 8y agoYes, that's how it is today. That's why I want proper merging and decay, so that fresh answers actually have a shot of getting to the top while people see a single relevant thread.
- mark-r 8y agoMaybe a good answer won't rise to the top, but it will go up over time. I answered a question 7 years after it was asked, and at this point my answer is #8 out of 59. I can't argue that the 7 above me aren't better.
- collyw 8y agoThere is also the problems of frameworks and languages evolving, with better ways of doing things in subsequent versions.
- Someone1234 8y agoInstead of decay, you could just embrace the concept of vote inflation. If a site continues to get more popular the number of votes (in both directions) will grow in proportion to that. It is normally a problem, but in this case could be an asset.
- Bartweiss 8y agoThis is an interesting point. For sites like Reddit, inflation is just a (minor) drawback because it means total vote count is a measure of (quality * recency) instead of quality. But for SO, where voting is never locked and content 'ages' at different rates, it's not so clear there's a downside. The "closed as duplicate" issue still stands, but rising user count is probably helping the problem.
- jessriedel 8y agoIt seems like the rate of site growth and the rate at which you expect answer value to age are independent, so it would be surprising if they were approximately the same and cancelled. And at the very least, SO popularity has to plateau at some point (if it hasn't already).
- EGreg 8y agoNo. Instead make the “Accepted” moniker have people do “Seconded” for whatever latest version it still works for. And then the newest users can come and see based on the LATEST version which is the most upvoted answer.
- manigandham 8y agoThat's just another "vote" then. What's the point of having answers with a vote and a second if people will just do both? The issue is that relevancy has a time-component, so there needs to be way to age scores as well to make sure the best answer for that time period is at the top.
- EGreg 8y agoIf there are no new versions, or the new versions have had a tiny uptake, then old answers are relevant!
- manigandham 8y agoSo what's the problem exactly?
- mixmastamyk 8y agoI don't mind the Accepted label, however they are sorted at the top by default. They probably should change that to "sort by votes," so the best answer can change over time.
- foxfired 8y agoAccepted answers are not necessary at the top. Here is an example: https://stackoverflow.com/questions/6121203/how-to-do-fade-in-and-fade-out-with-javascript-and-css https://stackoverflow.com/questions/6121203/how-to-do-fade-i... The chosen answer was not the best, so my answer ended up reaching the top. But note that after 7 years, my answer, although still correct, is no longer the most efficient way to do things. Better alternatives are rising up.
- mixmastamyk 8y agoIt's being sorted by votes. Many times, the accepted is at the top despite higher number of votes being cast for those below. I'm not sure what the deciding factor is now, perhaps there is a more complicated scoring system happening in the background or it recently changed?
- Flimm 8y agoThe only case where an accepted answer is not pinned to the top is when the accepted answer was posted by the same user who asked the question. Your link is an example of that.
- manigandham 8y agoWhat is the point of accepting an answer where there are multiple ways to do things, especially over time? It's a misleading signal.
- elboru 8y agoActually I've rarely considered duplicate questions as a bad thing, I like to open a bunch of related questions (usually a couple of them tagged as "duplicates") and check which one fits my problem better.
- irrational 8y agoRecently I have really started noticing this exact problem. Every answer (usually from closer to 2010) I have looked at is recommending using jQuery, even though there are perfectly fine ways of doing it without jQuery in 2018. I've taken to looking the dates of answers and ignoring any that date from more than 2-3 years in the past.
- ravenstine 8y agoLOL Half my SO searches include "-jquery".
- Someone1234 8y agoIf you still support IE11 very little has changed in the last 2-3 years. JQuery is still king. Companies that support large enterprise or government absolutely require IE11 support, since these places are required to use it due to exclusive e.g. Java Applets/Flash/ActiveX, etc. Microsoft's decision to drop those technologies from Edge was kind a mixed bag. From a security perspective it was the right decision. But it has also stuck many of us developing for a non-evergreen browser, even while our users are on Windows 10. IE11 is the new IE6.
- rc_kas 8y agoYeah this is becoming a real problem. I lookup an error and find the solution solved with an older version of the framework. If I'm lucky someone left a comment regarding the version I'm on, but not often.
- crdoconnor 8y agolol I find it amusing that the highest and lowest ranked comments on this page made the exact same point.
- beached_whale 8y agoThere is a long tail to tech use and the answers often are the best for that particular set of variables. I think that adding tags to make it more specific in this regard as time passes can really help. If it stops being relevant when library X reached version Y, add a tag that says pre-version Y
- BeefySwain 8y agoI started learning Python this year. The number of Python 2 specific answers on Stack Overflow has been a not-insignificant barrier to that. Almost all answers recommend using os rather than pathlib, many answers recommend using cmp functions rather than keys for sorting, etc. Not to mention all the print statements.
- mapgrep 8y agoWhen he writes about ohw StackOverflow is first and foremost a wiki, he is offering a solution to this situation. In the original vision of StackOverflow, that accepted and highly voted answer would be converted into a wiki. Then when it gets stale, it would be revised. Not saying they do a good/bad job of this, just that, that's how it was supposed to work. And ya it would be totally frustrating as a new contrib because if the canonical answer was not converted to wiki yet, you can't really contribute.
- mark-r 8y agoThe idea of StackOverflow as a wiki is completely at odds with its built-in gamification. Answers are written by individuals who garner points for them, and their name and avatar are prominently displayed.
- ant6n 8y agoA related issue: the best answer may have a link to the documentation, which may rot. Now people will down-vote the answer because broken link. This could happen if somebody asks how to do something using some library. If the library actually has a builtin facility to solve the problem, but then the answer gets down-voted due to documentation link rot; then another answer explaining how to built that facility from scratch may become the 'best' answer, even though the re-implementation of existing library functionality is a bad idea.
- mark-r 8y agoThe best answers paraphrase what's said at the link, so they don't become worthless when the link goes stale.
- geezerjay 8y ago> One challenge not discussed here is when an accepted and highly voted answer _used_ to be a good answer, but no longer is in later versions of whatever software/platform/library is discussed. That's not a big problem. Questions and comments can be edited by other users, which includes adding and altering tags that reflect software versions.
- InclinedPlane 8y agoEditing someone else's question or answer (comments can't be edited by others) is a very aggressive move and not something that anyone other than a handful of very experienced / high rep users are likely to do very often. And it is far from a perfect solution either. Editing an answer substantially upsets the entire up/down vote system, as well as the comments. A better system would be something like question/answer versioning. In general it's a problem that has received far too little effort.
- geezerjay 8y ago> Editing someone else's question or answer (comments can't be edited by others) is a very aggressive move and not something that anyone other than a handful of very experienced / high rep users are likely to do very often. Do you really believe that adding a tag with a version number to a post is something that can be classified as a very aggressive move, particilarly if it's an old post?
- InclinedPlane 8y agoPlease be serious, you know that's not what is meant by editing old answers in the context of "stack overflow is more like wikipedia than anything else". The implication is wholesale change of the entire content of the answer (or question). Which fundamentally breaks how stack overflow works, and is why it's not something that happens very often.
- thaumasiotes 8y agoThat's not a solution. Tags are hidden at the bottom of the answer. Accepted status and vote count are prominently highlighted. Accepted status locks the outdated answer to the top of the page regardless of whether it's valid.
- teh_klev 8y agoThis is a pretty good example of this problem which got a top of front page hit for my google search (and it's been viewed 15k+ times): https://stackoverflow.com/questions/36747293/how-to-specify-the-view-location-in-asp-net-core-mvc-when-using-custom-locations/36772778 https://stackoverflow.com/questions/36747293/how-to-specify-... The question is tagged for ASP.NET Core 1.0 so there's answers that are applicable to that version. Then things changed in 2.1, which removes some friction when adding custom view locations, and another user posted an answer specifically for that version.
- btilly 8y agoAnd this can happen surprisingly quickly. As a cautionary example, see https://stackoverflow.com/questions/5021456/how-does-mongodb-avoid-the-sql-injection-mess/5021598#5021598 https://stackoverflow.com/questions/5021456/how-does-mongodb.... The question was why MongoDB was by default able to avoid the equivalent of SQL injection attacks. My answer took what was in their FAQ at the time, and expanded upon it based on advice from D.J. Bernstein - who is famous for his ability to write secure by default code. Then it was discovered that PHP volunteered to create security holes in MongoDB. Then it was discovered that multiple other languages and libraries likewise volunteered to do so. Had they followed the good advice that I was quoting, the problem wouldn't have happened. Despite the other faults of MongoDB itself, in this case it was truly innocent. The security problem wasn't in it, but in the libraries that were written around it. It made it easy to write secure code and they didn't. But the result? From the point of view of a naive programmer, not long after my answer appeared it was revealed as misleading for PHP programmers. A few years later, it was actively misleading for other languages. And while the advice remained technically correct - which is why I left it - it has been criticized on reasonable grounds by several other people since.
- codinghorror 8y agowhy haven't you, or someone else, edited the answer to indicate this or add a disclaimer?
- btilly 8y agoWhat makes you think that I haven't? For a long time I thought that the comment that I left in 2011 after the PHP bug was sufficient. Remembering it today I edited the answer at the same time that I posted here. Now I could be faulted for not having responded to the comments in 2015. However I was taking an extended break from stack overflow after getting annoyed at the moderation of https://stackoverflow.com/questions/11314077/algorithm-for-exclusion-of-numbers/11317787 https://stackoverflow.com/questions/11314077/algorithm-for-e.... When I came back, dealing with the variety of things that could have further annoyed me while I was gone did not seem worthwhile, so I didn't. (It remains a pet peeve that a lot of moderators remain quick to close algorithm questions because they do not recognize DP-type problems as things which are more likely to be answerable by people with a CS background than a math one.)
- PeterisP 8y agoThis is a major issue. Just last week I had a student make an untrue claim about how something works based on a top ranking Stackoverflow answer. It turns out that the answer is from 2008, and was true back then, but has not been for something like 5 years already. Stackoverflow badly needs a system for people (moderators?) to be able to 'obsolete' answers, possibly mark them with a highly visible "valid only up to version x.y of 2010" tag or something.
- codinghorror 8y agoTry pressing the "improve this answer" aka edit button and editing it, for starters. Works even for anonymous accounts!
- codinghorror 8y ago> One challenge not discussed here is when an accepted and highly voted answer _used_ to be a good answer, but no longer is in later versions of whatever software/platform/library is discussed. This is also something that gets worse over time, like the uniqueness problem. It's not discussed because the solution is kinda obvious.. wikis mean editing, right? Even by anonymous users? That's also how it works on Stack Overflow. Edit the answer to make it better, more up to date, or add a disclaimer at the top that it only applies to very specific versions as needed.
- Izkata 8y ago> Edit the answer to make it better, more up to date This is strongly discouraged because then the existing votes become a lie. > or add a disclaimer at the top that it only applies to very specific versions as needed This I think is allowed, but is extremely rare.
- codinghorror 8y agoEditing is not strongly discouraged, and adding a disclaimer is not rare.
- Izkata 8y agoEditing in general no, changing the answer is. You get rebuked and told to post your own.
- explainplease 8y agoYep, and then the better answer will sit at the bottom of the page, while the older one continues to be seen and gathers more upvotes. I think there is a critical mass after which the site's focus changes from creation to protectionism, and then it becomes stagnant.
- jrochkind1 8y agoThe solution to dups is obvious too, right? Close em as dups? Yet that is a large focus of the article. The solution to out-dated answers might be obvious, but apparently obvious isn't stopping it from being a problem... it's a big problem. If it's obvious, why is it still a problem? If the answer is "because the userbase isn't doing what I want", then _that's_ a problem to which the answer is apparently not obvious. To the extent that SO has been succesful, it is of course that figuring out how to structure the software so as to get the userbase to do useful things _is_ the problem, not an "exersize left to the reader".