What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 128 Stack Exchange communities.

Something's been bugging me about SO for a while now, that is the inaccuracy of "length of time" strings. The best example I can give at the moment is CMS's user profile. If you look at the length of time he's been registered, it says 1 year. He joined the site September 2008, which is only 3 months off two years.

It's the same with answers - you can give an answer 50 seconds before anyone else, but at certain times page views will show both as "5 minutes ago" or whatever. That's not quite as bad, but with answers being sorted "randomly by votes" by default you could lose out on votes or being marked as the correct answer to the slower answer even if they're both the same.

There's probably a better method amongst these, IMO:

  1. "Vulgar fraction" unicode characters
    • Member for 1½ years
    • answered 5¼ minutes ago
    • asked 3⅓ hours ago
  2. Different units
    • Member for 21 months (at least it's better than "1 year")
    • Answered 306 seconds ago (ugly, I know)
    • Asked 64 minutes ago (also ugly)
  3. Combining units
    • Member for 1 year 9 months (fits easily on the profile screen)
    • Answered 5 minutes 6 seconds ago (probably too long)
    • Asked 3 hours 20 minutes ago (also long)
  4. Rounding up

I personally prefer option 1, unicode fraction characters. "Member for 21 months" is pretty good too because it's easy to quickly work out. "Asked 452 seconds" ago is horrible though.


EDIT: I'm completely aware of the timestamp hover because I am forced into using it so much. It's a great feature, but it's not really in your face so not everyone knows about it. I still think there's a better "glancing" solution than having to hover over the time string.

share|improve this question
    
Jeff was against changing "yesterday" to an actual date/time. Good luck with this one. –  Jon Seigel Jun 14 '10 at 23:41
    
@Jon: yeah. I can see why a string like "yesterday" is desirable. I currently workaround time format issues on answers by having "oldest" as my default sort. It just seems like it could be improved. –  Andy E Jun 15 '10 at 7:47

2 Answers 2

If you hover over any time display, you get a tooltip for the exact timestamp, down to the second. For example, CMS registered at 16:04:18 on September 9th, 2008. This applies to anywhere that a condensed time display is shown on the site (besides the timeline view, which is only partially completed anyway (which why there is no way to access it through the UI)).

This utility is very accessible, and I don't think the extra utility in expanding the condensed time display will help too much. Rounding up values is inaccurate (CMS hasn't been around for 2 years, for example). Subsequently, fractions really break up the stream of text, and you already cover the issues with options 2 and 3. In the end, I think the current system gives a fine balance between accuracy and still looking good.

share|improve this answer
    
-1 (no offense intended). I'm fully aware of the timestamp on hover, because I'm forced to use it so much. That is how I was able to put the date that CMS registered in my question. Although it's accessible I disagree that everyone knows about it. I also disagree that "Rounding up values is inaccurate", because in the case of CMS he's far closer to being registered for 2 years than he is for 1 year, and the current system gives a "fine balance between accuracy and still looking good" is just wrong. Sure it looks good, but it's not at all accurate. I still think there's a better solution. –  Andy E Jun 15 '10 at 7:22
    
@andy your question is ... poorly asked IMO. You're really proposing a slight change to the user page "member since" formatting (which I actually agree with) but it comes off as this way broad, nebulous request. –  Jeff Atwood Jun 15 '10 at 7:44
    
@Jeff: I don't think this problem just applies to the "member since" section. I don't think "3 hours ago" is accurate enough when one user posted 3 hours ago and another posted 3½ hours ago. It results in a random selection between choosing the answer, even if both are exactly the same. –  Andy E Jun 15 '10 at 8:49
    
@Andy We're operating on two separate ideals of accuracy. You're operating under the "closest to the actual number", I'm operating under "A value that is correct". Saying CMS has been around for 2 years is technically a lie when he hasn't been around that long, so it would be an inaccurate statement. I'm not saying that your measure of accuracy is wrong, but I don't believe my measure is wrong, either. –  Grace Note Jun 15 '10 at 9:50
1  
Also, no offense taken. I may look young into the system, but it'll take a lot more than a measly -2 points to do me harm. Besides, you had the courtesy to explain the downvote, anyway, so at least I don't have to infer some generic "Your answer didn't work" like with other downvotes. –  Grace Note Jun 15 '10 at 9:55
    
@ccomet: I agree, although I think neither rounding nor the current implementation is anywhere near perfect. I keep looking at my comment above, where I typed "one user posted 3 hours ago and another posted 3½ hours ago", and I think "that looks pretty good". I'm not saying we should go to extremes like ¼, ⅓ or ⅜ - but ½ looks fine to me. –  Andy E Jun 15 '10 at 11:48
    
@Andy I wouldn't call it perfect either, but it satisfies my need for accuracy. One issue I might think with using fractions is that it will require us to track half-units for all units of time. Which is trivial mechanically, but I don't see strong benefits to updates within the half-minute or half-second marks, nor any real reason to track that I, for example, been around for 60 and a half days. So seeing "user posted 43½ seconds ago", to me that looks ugly and useless. That may just be me, though. Only hours for posts and years for users is really any use to expand like that... (cont.) –  Grace Note Jun 15 '10 at 12:12
    
...because they're the only periods of time that a half-unit really has a strong impact. The difference half-units of the other times make is just insignificant. Only applying the half units to those two time cases would be inconsistent, which to me would likewise be ugly UI. Again, maybe it is just me that this is a lose-lose situation for implementation, but it just doesn't look right now matter how I arrange it. –  Grace Note Jun 15 '10 at 12:17
    
@ccomet: I see your points. I think ½-minutes are necessary for the reasons I said to Jeff - random sorting can cost you votes where the voter cannot see the time difference at a glance. I certainly wouldn't expect ½-seconds to be implemented. 3½ days, 2½ hours, 1½ minutes, 12 seconds... I don't think it looks that bad. Maybe it's you, maybe it's me! ;-) –  Andy E Jun 15 '10 at 13:17
    
@Andy You seem to be conflating the time we see with the time the system knows. The system doesn't confuse two posts that both read "2 hours ago" as being posted at the same time -- it knows exactly what time they were posted at. As for answer shuffling, it happens for posts with the same votes posted within 10 minutes of each other; an answer posted 11 minutes after another will be stuck below it no matter what timestamp they display –  Michael Mrozek Jun 15 '10 at 13:58
    
@Michael, I'm not confusing that at all, I think you're confusing what I wrote ;-) I just wasn't aware that it wasn't shuffled if there's more than 10 minutes difference. –  Andy E Jun 15 '10 at 14:08
    
@Michael There is no such mechanism as far as I can tell. I tested this in a question I know with 4 answers, and I did get a combination where an answer was listed higher than an answer given 20 minutes before. Where did you find this anti-shuffling notice? –  Grace Note Jun 15 '10 at 14:40
    
@ccomet Oh, maybe I made that up. I was thinking of this, which I remember reading at some point, but that was never actually accepted –  Michael Mrozek Jun 15 '10 at 15:25
    
@Andy See, I disagree that the caused cost of votes is a problem, nevermind one of any significance when you dive into fractions of a minute. Without such a display, the window in which two questions will be shown with the same minute despite differing by over half a minute is less than 30 seconds per minute. Which is a pretty good bias in favour of the earlier question. And when you look at the points where answers are less than half a minute apart... it's a competition to who just hit the "Post your Answer" button first, not who got there to help first. –  Grace Note Jun 15 '10 at 16:51
    
Thinking on it now, the whole point of randomizing the order was pretty much to try and reduce the emphasis that time difference had on voting. I'm not naive enough to think that nobody votes based on time, but I think that putting strong emphasis on tiny units of time is a step in the wrong direction. People are already made to try and view all of the answers, so if after that they still want to judge on something like post time, then I think that some effort on their part makes sense. It's a metric, yes, but not one I think should be encouraged by the system. –  Grace Note Jun 15 '10 at 17:05

ccomet's answer is what you're looking for, but you also might find the timeline view helpful for determining what's happened on a particular question

Edit: You also might be interested in the original blog about this, which I just happened to run across while looking for something else in the archives

share|improve this answer
    
ccomet's answer isn't what I'm looking for, really :-) I'm aware of the timeline, but I'm not sure how to access it without typing it into the address bar. It doesn't seem very accessible at all. –  Andy E Jun 15 '10 at 7:27
    
@Andy The timeline view is still "experimental", last I heard. I'm not sure when they plan to add a link to it –  Michael Mrozek Jun 15 '10 at 13:58

You must log in to answer this question.

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