Back in the days before being able to comment on posts (circa September 6th 2008 according to this blog entry), post id's < 35301), rightly or wrongly answers were often used for communication between users.

In my moderation duties I frequently see "not an answer" flags raised against some of these old fossils from some of our more tenacious users and Stack Overflow archaeologists.

Many of these old posts can be in a fair old mess and are hard to clean up because:

  • some of these non-answers have @user but over the years folks have changed their display names so it's no longer possible to identify who the "comment" was intended for.

  • converting these answers to comments will cause the user who's post received the converted comment to get an inbox alert. This isn't ideal, because of the pre-cambrian nature of the whole Q&A thread, it's a bit of a red-herring and probably annoying.

  • many of these old "answers" are often a bit hybrid and still carry some useful information

This was an example of an answer that got flagged:

How do you kill all current connections to a SQL Server 2005 database?

(Ok, as others have pointed out I didn't spot that this user also had an accepted answer as well and so there could have been an answer merge there)

I decided to leave it as-is because it's both a comment and an answer, and converting to a comment would probably annoy "Adam" but also its time has long passed.

I'm proposing that if one of these Archean answers/comments (posted with good intentions at the time) is spotted in the wild then maybe we should just let it fossilise?

Obviously if it's just junk and clearly of no good use then it needs to be removed, but I'm thinking that the more genuine answers of this type be left alone because of the reasons listed above.

What does the community think?

  • In the specific case, the user who wrote the answer you report left two answers that could be merged together. – kiamlaluno Jul 13 '11 at 14:58
  • 1
    @kiamlaluno - Probably, but when you're moderating you tend to deal just with the flag that's been raised. On longer posts such as that it takes a while to assimilate all that info and sometimes you just see a big wall of text, especially when there are more pressing flags to deal with. – Kev Jul 13 '11 at 15:01
  • I agree with that. I am saying that the answer, rather than be flagged as a post, should have been probably flagged for another reason, as the one I reported. Clearly, moderators focus on the current flag, but who flagged could have noticed there are two answers from the same user. – kiamlaluno Jul 13 '11 at 15:09
  • For what it's worth I went ahead and merged the two answers now with an edit, making the second one redundant. – Tim Stone Jul 13 '11 at 15:32
  • possible duplicate of Please discourage flags on old questions and answers – Cody Gray Jul 14 '11 at 3:57

I'll admit to being someone who occasionally goes actively hunting for non-answers via search tools, which tends to be a date-agnostic activity, so I'm probably partially responsible for some of those "streams of flags". I've certainly seen a lot of these borderline cases you describe, and my rough procedure is to flag answers whose only content falls into one or more of the following categories:

  • Requests for pre-existing information:
    • e.g., "what version of {...}are you using?", "could you show the definition of the {...} function as well?", etc.
    • but not requests for diagnostic information, e.g. "what happens if you try doing {...}?"
    • justification: Not relevant after the fact. Gives no information about a solution or how to find one.
  • Uninformative confirmations of reproducibility:
    • e.g., "I get the same result when I do {something equivalent under equivalent circumstances}".
    • but not references or informative confirmations, e.g. "this looks like bug #4922 on their issue tracker", "this happened to me when I upgraded from version {...} to {...}"
    • justification: Most of these are very situational, and provide little value other than (perhaps) confirming that the issue is not something peculiar to the OP.
  • Asking about solutions:
    • e.g., "@{...}, could you please clarify the third step in your answer?", "what did you end up doing about {...}?", "I'm having the same problem, please help",
    • justification: Completely useless as an answer and very unlikely to produce useful results.

In contrast, I try not to flag anything that might have information for someone arriving from a web search, even if it's only a suggestion of troubleshooting techniques, references to other information, speculative starting points for a solution, etc. I'm sure I'm not completely consistent, and likely to err on the side of flagging too much, but the above is what I aim for.

If this sort of thing isn't actually useful, I'd love to hear about it so I know what to stop doing. I want to make SO better, not create useless work for moderators.

  • 1
    I use similar guidelines when flagging old answers, essentially if it doesn't have anything constituting an answer to the question, it gets flagged. If it contains a solution or an update to a solution, I usually leave it. – staticbeast Jul 13 '11 at 18:35
  • 1
    Pretty much, yeah. What it mostly comes down to is that if an answer would be useful as a comment or edit, I'll leave it alone if it's old. If it's completely useless, I'll flag it. – McCannot Jul 13 '11 at 18:38
  • +1. I identify with your last sentence especially; I also search for flaggable posts, and would be grateful for more enlightenment. I can't imagine it's that bad, since all of my flags have been marked Valid; I'm sure they would personally contact us if we were doing something wrong. – Rob Hruska Jul 14 '11 at 2:30

If the "answer" is clearly not an answer, and is not a hybrid answer/comment, or is along the lines of, "hey, guy, I couldn't get the code in your answer to work. plz send?", then flag it as "not an answer" with the intention of it being deleted.

If the "answer" is a hybrid answer/comment and would be easily migrated to a comment or merged, then the burden should be on the flagger to find out as much as they can about the context of the "comment/answer", and provide that information in the "notes" textarea when flagging, using the "other" reason.

The information the flagger should provide might include any of:

  • If the "answer" belongs as a comment on another answer, along with a link or note to which answer it should be migrated to.
  • If the "answer" should be merged with another answer, along with a link or note to which answer it should be migrated to.

However, if all of the following are true:

  • The "answer" is too extensive to be migrated to a comment
  • The "answer" cannot be merged into another answer
  • The "answer" contains valuable information that might help the OP or other viewers of the question.

Then it should be left as is, and not flagged.

  • I agree, but migrating to a comment is icky because (as I mentioned) the user who's post (we can convert to a question or answer comment) receives the comment will get an inbox alert. Maybe folks do want to be alerted to changes in these old Q&A's like this, I don't know. – Kev Jul 13 '11 at 15:04
  • 2
    If it were me receiving it, I wouldn't care too much about the comment notification. Maybe the notification is actually good? If the original answerer never checked back on the question, they may not have known that the commenter posted an "answer" in reply to their answer (since there's no notification when other users post answers to a question you answered). – Rob Hruska Jul 13 '11 at 15:15
  • 1
    For what it's worth, I haven't always adhered to this "policy" in the past, and I do flag a lot of stuff. However, I'll try to provide more information in the future when flagging posts like this. – Rob Hruska Jul 13 '11 at 16:00

What has been posted as an answer, but it is not an answer, it still is not an answer, independently if the answer has been written two days ago, one month ago, or one year ago.

I agree that, if users start to flag old posts in a massive way, that would make handling the flags more problematic, but considering that such flags are visible to 10k users, it's not just the moderators that handle those flags.
Clearly, the priority should go to new posts, but that doesn't mean that old posts should be completely forgot.

Also, if the user goes hunting for posts to flag, that is probably something we don't desire it happens; if the user finds a post that thinks should be flagged while searching for a question, then that is different, and it doesn't change basing on how old the post is.

  • 2
    I don't think @Kev is saying stop flagging, just don't go looking for stuff to flag. – ChrisF Jul 13 '11 at 15:10
  • 1
    I agree and I added some caveats to my case. However there are days when you get a stream of these flags and these questions have accreted all sorts of stuff which can be hard to untangle, takes a fair bit of time, and meanwhile there are 200+ more pressing flags on the moderation queue. I only suggesting that some common sense comes into play...or of course folks with editing rights help to clean up these posts (merging multiple answers etc) then flagging the stuff that then needs removing. – Kev Jul 13 '11 at 15:13
  • I didn't say that hunting for posts to flag was something we want to see happens, but I made that explicit in my answer. I read the question as regarding old post being flagged (there is no evidence the user who flagged the post was hunting for posts to flag), but I could read it wrongly. – kiamlaluno Jul 13 '11 at 15:19
  • 3
    @Kev, I usually try and do the research and tell the moderator the complete solution if there is one (and do any editing I can do myself). In that situation, I would have edited the accepted answer first, then flagged the moderator to delete the unnecessary one. – Lance Roberts Jul 13 '11 at 15:20
  • @Kev I agree with you about the common sense. I would not like to see 200 old posts being flagged in a day, together with other 200 new posts (supposing that 400 is the higher number of flags done in a day). – kiamlaluno Jul 13 '11 at 15:22

I agree with your general concept, but that answer should be merged with the accepted answer by the same author, since it's just a follow-up to it. No one need be notified in that case.

  • It was late and I missed that, there's a lot of answers there and my attention-fu was a bit low. – Kev Jul 13 '11 at 15:07
  • @Kev, know what you mean, that's why I think I have the record for editing my own edits. – Lance Roberts Jul 13 '11 at 15:13

I'm thinking that if it is an answer, however communicative it may be in nature, then it should be fine to leave around. This may be more often the case with older answers than with the more recent not-an-answer junk.

However, I think that we have a good responsibility to be consistent and retroactive with policies like this. Yes, it was excuseable back then. But if there's one thing an insistence of keeping things "for old time's sake" has demonstrated, is that a lot of people will throw a beef about their behavior being the same as before. Having the explanation is just an excuse, it doesn't reduce the frequency that it happens.

It's certainly not as bad-behavior-promoting as other examples, but I think that it produces an overall cleaner site if any non-answers are handled, old or not. It shows better that we do enforce the rule.

On the subject of the inbox. If you get an event in your inbox, it is assigned in accordance to the time of the event, not the time it shows up in your inbox. When you convert an answer into a comment, it is assigned as having been posted at the time of the answer's posting.

As such, you will still ping the user with a number on their inbox. It will, however, most likely be a ghost-ping that will leave them scratching their head since the corresponding message will not exist. If, that is, they don't have other elements highlighted that they might consider just being more recent old news.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .