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.

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.

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.

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?

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.

And then the newest users can come and see based on the LATEST version which is the most upvoted answer.

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.

Huge swathes of features appear at transition points like MySQL 5.5 -> 5.6, which means lots of answers are now viable but no longer the best option. And since they are still correct, it's hard to write up new questions which won't be closed: instead of saying "Answer X no longer works", users have to say "Is Answer X still the best approach?" and get closed for duplication or subjectivity.

Worse still, lots of databases still have multiple versions in active use, so "worked great for me! - posted yesterday" isn't even a helpful reference.

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.

I think Microsoft has made a mistake by stopping the addition of standards-compliant features to IE11. They should keep supporting it and modernizing it until the userbase has moved to Win10 on its merits, instead of trying to use IE11's aging as a lever to force Win10 adoption.

I do think a mistake was that IE11 was included at all in Windows 10 to appease corporations, instead of just ripping off that particular band-aid. But doubling down on it and allowing corporations to default to nothing but IE11 via group-policy is the biggest mistake. GPO footguns like that are why corporate America can't have nice things, and I'm increasingly of the opinion that it's time Microsoft started dropping footguns from GPO.

(Also, note that the root problem in this particular workplace remains Oracle and their web apps that are stuck in IE6 worst practices but cost millions of dollars each year to continue to operate. Oracle is the real villain here.)

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.

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.

If I'm lucky someone left a comment regarding the version I'm on, but not often.

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.

I'll often find answers to questions which are outdated and haven't been improved. I agree they have a mechanism in place to fix it, but are users actually fixing answers en masse? SO could introduce an age out mechanism that requires an answer to be maintained semi regularly? Do they already have this today?

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.

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?

Yes, it really is. Tags are used to filter results. Stackoverflow specifically lets users run searches only on specific tags, or even exclude certain tags from the search results. As tags reflect technologies and their versions, stackoverflow enables users to run queries only on specific versions.

99% of people use Google as the search interface to stack overflow.

As someone who primarily interacts with SO via search engine results, it's one of the biggest problems with SO. The number of top voted/accepted answers that are 'wrong' in one way or another keeps going up and thus the chance of me finding the 'right' answer on SO keeps going down.

Otherwise, how is it "not a big problem"? Is it the sort of thing where "in theory" it's not a big problem, even though in actuality it significantly effects the usefulness of the site to many people?

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.

My edit was quickly rejected by a user who had no knowledge whatsoever of the technical domain in question, claiming that my edit went against the spirit or intention of the original answer's author. I then submitted mine as a separate answer, where it languishes at the bottom of the page, rarely noticed by users who come looking for the answer (and it's virtually the only hit on Google for the question, has been for years). They all see the outdated, poorly organized one instead.

So my work was a waste of time, shot down by an ignorant, mindless policy wonk whose only goal is to farm reputation points.

I have not contributed to SO since. I have better things to do with my time, like publish my own documentation on readmes on GitHub.

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".

When SO was new, there was incentive for people with expertise to contribute it. Now there is more incentive for Wikipedia-types to go around smashing others' contributions with their rejection hammers. It's a waste of time for those who actually have something to contribute.

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.

I think there is a critical mass after which the site's focus changes from creation to protectionism, and then it becomes stagnant.

As a cautionary example, see 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.

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-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.)

It is amazing the number of replies in this entire HN thread that have no idea every single question and answer on Stack Overflow is editable, even by an anonymous user. The first item in my list was Stack = wiki for a reason!

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.

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.

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.