25
$\begingroup$

Recently I noticed that the software inserts the id of the currently logged in user in the “share” link of any MathOverflow question or answer. When one is logged out, the “share” link comes out normal, without any private information. For example, the same link appears as https://mathoverflow.net/a/198099/402 or https://mathoverflow.net/a/198099 when I'm logged in/out, and 402 is my id on MathOverflow.

I think it's safe to assume that users have a reasonable expectation of privacy with respect to links to publicly available pages, in particular, one does not expect one's own unique user id to be inserted in the link, allowing others to identify the user that posted the link.

One doesn't have to go too far to see that this type of surveillance activity is far from being innocent: an anonymous referee reviewing a paper for a journal adds a link to a MathOverflow answer in his report, e.g., in order to clarify something. Unbeknown to him, the SE software inserts his user id in the link, revealing his identity to the author.

What is the opinion of the MathOverflow community on this matter? Do we consider such type of behavior by SE to be acceptable?

Myself, I see it as an unethical invasion of privacy, especially because the users are not informed of this type of surveillance.

$\endgroup$
  • 7
    $\begingroup$ I removed the bug is it is certainly "by design." That you do not like it is your right, but it is clearly not a bug. $\endgroup$ – user9072 Mar 8 '15 at 22:41
  • 5
    $\begingroup$ One student of mine from last semester sent me her personal password for the university website. In retrospect she said "Oh well, what's the worst that could happen?", and I replied that the personal information include, amongst other things, banking information and other delicate information. And that she should never do that. If someone doesn't understand the tools they are using, they might make mistakes. This applies here as well, if the referee gives out their email address, by accident or without knowing it will reveal their identity, it's still their fault. $\endgroup$ – Asaf Karagila Mar 8 '15 at 23:00
  • 12
    $\begingroup$ @AsafKaragila: The important distinction between these two situations that you fail to make is that the reviewer is aware that giving out their email address will reveal their identity, whereas the same cannot be said about a link to public web page, and as I mentioned in the original post, one has a reasonable expectation of privacy with respect to links to public web pages. Thus revealing one's email is a negligent behavior whereas revealing a link to a public web page is not. $\endgroup$ – Dmitri Pavlov Mar 8 '15 at 23:25
  • 2
    $\begingroup$ @DmitriPavlov an email-address is perhaps an extreme example, but various programs add meta-information to files that can be identifying. There are some journals that account for this, but by no means all. And, again, whoever gets that URL clicked a share link, with all kinds of social-media icons below. If they want to use it in a scientific context why not click "cite" there. Problem solved. $\endgroup$ – user9072 Mar 8 '15 at 23:38
  • 1
    $\begingroup$ @quid: I disagree, often only a specific answer is needed, but in any case this is not for us to decide. $\endgroup$ – Dmitri Pavlov Mar 9 '15 at 0:07
  • 1
    $\begingroup$ @quid: It is not for us (i.e., you, me, and the users of MO) to decide how the referee elects to cite a MathOverflow answer in his report: as a direct link or as a BibTeX entry provided by “cite”. $\endgroup$ – Dmitri Pavlov Mar 9 '15 at 0:37
  • 2
    $\begingroup$ @DmitriPavlov if they want a direct link they can just copy the URL from their browser's navigation bar. They would choose to share a link, for this to happen what you are worried about. Look, I saw people post links to journals that did not work since they went through a proxy of their institution and this did reveal the institution; I think even on this site. Did an institution behave unethically for setting things up in this way? $\endgroup$ – user9072 Mar 9 '15 at 0:42
  • 3
    $\begingroup$ @quid: I think we're going in circles here. Please correct me if I'm wrong, but the only way to get a link to an answer is by using the “share” link. Copying the URL of the page itself only gives a link to the question, not the answer. Finally, I don't see how the word “share” would imply that the resulting link is under surveillance. $\endgroup$ – Dmitri Pavlov Mar 9 '15 at 11:55
  • 1
    $\begingroup$ @quid: As for your example with proxy servers, the answer depends on whether or not the university name is clearly discernible in the URL. $\endgroup$ – Dmitri Pavlov Mar 9 '15 at 11:57
  • 1
    $\begingroup$ "Please correct me if I'm wrong" You are welcome. Please consider for the future to read resources provided to you before asking follow up question. This might also make it easier to make progress in the discussion. (Sidenote on the linked answer: There are other ways to do this too.) $\endgroup$ – user9072 Mar 9 '15 at 13:20
  • 1
    $\begingroup$ I have to wonder. Do you raise this point because you know someone whose referee used such link; do you know a referee that used such link; or is this just a hypothetical? $\endgroup$ – Asaf Karagila Mar 9 '15 at 16:30
  • 1
    $\begingroup$ I didn't ask about leaking information. I asked about this particular instance. $\endgroup$ – Asaf Karagila Mar 9 '15 at 17:08
  • 1
    $\begingroup$ Should we create poll-type answers to get a feel for whether people care or not about this issue, and whether they have a positive or negative opinion on the matter? $\endgroup$ – Ricardo Andrade Mar 9 '15 at 22:06
  • 3
    $\begingroup$ @RicardoAndrade first, it should be clarified what "this issue" is, as OP is not quite clear. for example is it: a) that track-able links exist at all or b) that it is not clear enough that they exist c) that it is not clear enough when a link is a track-able link d) that there is not more convenient access to a nontrack-able one. Still something else? Here is a concrete proposal: .) I do not consider the current situation as problematic. .) I wish the feature of track-able links remove entirely. .) I wish it is made more clear when a link is track-able and easy access to non-ta version. $\endgroup$ – user9072 Mar 9 '15 at 22:37
  • 1
    $\begingroup$ @quid, you are right, I was not being specific. I think your three suggestions for poll-type answers are a pretty good start. Perhaps the OP will chime in with his opinion. $\endgroup$ – Ricardo Andrade Mar 9 '15 at 22:41
38
$\begingroup$

I consider the current situation as problematic, and would like to see this problem adressed, e.g., by making the nature of the link more transparent and/or providing direct-link without UserID in addition with the same visibility.


This is a poll answer. Please do not downvote. Up-vote competing answers instead.

$\endgroup$
  • 1
    $\begingroup$ I am voting for this option with two understandings: first, while the situation is problematic, the problems it presents are likely to arise quite rarely and are usually not too serious. Second, I would want the link without UserID to have greater visibility, not just equal. $\endgroup$ – Eric Wofsey Mar 11 '15 at 5:09
  • 1
    $\begingroup$ I don't agree this is a problem (and have voted on the other poll answer), however, a potential solution could be to allow users who do care to opt out in their profile options. Their user ID would no longer be included in links, with the cost being the inability to obtain Announcer badges and related features. It seems rather unwieldy (and confusing to explain to new users / not readily apparent) to try and work direct links into the web interface along side the current system. Other solutions seem overcomplicated.... $\endgroup$ – Jason C Mar 25 '15 at 17:13
  • $\begingroup$ @JasonC: I don't see how that would address the issue at all. As mentioned in other comments, there are not-too-difficult ways to obtain anonymous links already; the problem is that many if not most users are not even aware that the "share" links are not anonymous, and that the non-anonymous links only exist for a "feature" that many of us consider to be of dubious value (especially on a site like MO). If including both an anonymous and a non-anonymous link is unwieldy, it seems to me that the obvious solution is just to get rid of the non-anonymous link. $\endgroup$ – Eric Wofsey Mar 26 '15 at 18:37
  • $\begingroup$ @EricWofsey what if it were opt in not opt out, that is one could switch-on the personalization in the profile? $\endgroup$ – user9072 Mar 26 '15 at 18:42
  • 1
    $\begingroup$ @quid: I would have no problem at all if it were opt in. $\endgroup$ – Eric Wofsey Mar 26 '15 at 18:43
  • $\begingroup$ @EricWofsey Opt-in works too but then you have the opposite problem; the same users who wouldn't know about opt-out wouldn't know about opt-in and so you lose some site features. One way to present the opt-in/out that could solve this is to show a notice to existing users (if the change is made) and provide a notice to new users to make them aware. $\endgroup$ – Jason C Mar 26 '15 at 19:05
  • 1
    $\begingroup$ @JasonC: Of course, but in this case the feature is one that many MO users wouldn't mind losing and some would even like to lose; see this previous discussion. $\endgroup$ – Eric Wofsey Mar 26 '15 at 19:28
  • $\begingroup$ @EricWofsey If the default opt-in / opt-out could be made settable per site (and if, on roll out, that per-site default was applied to all existing users, with notifications sent out as well), that could be a good way to let each site choose its own behavior. MO's default, then, could be to disable initially and require opt-in. $\endgroup$ – Jason C Mar 26 '15 at 19:36
17
$\begingroup$

I consider the current situation as satisfactory and do not see much need for a change.


This is a poll answer. Please do not downvote. Up-vote competing answers instead.

$\endgroup$
16
$\begingroup$

Two points that I haven't seen highlighted enough in the rest of the discussion.

  1. The only (apparent) purpose of the userids in the permalinks is awarding Announcer badges. Badges and reputation may be fun, but they are just a game and they are not that important. If this is the only reason in favor of uids in permalinks, I think that the cons highlighted by @DmitriPavlov in the question (there is a plausible scenario in which they can reveal a referee's identity) outweigh the pros. If they are used internally for other "surveillance" purposes (which I find unlikely), this tilts the scale even more in the same direction.

  2. It is not necessary to make the sharer identity public in order to award these badges. The problem can be easily addressed with some crypto. Just take the userid, salt it, encrypt it with a symmetric key, and throw the result in the permalink. Or, alternatively, use the same method as URL shorteners such as bit.ly: every new click on "share" creates a new entry in a database containing (random_URL_string, target_question, sharer_uid); Stack Exchange will keep this database private, so they know who created each permalink, but the public doesn't.

Considering these two facts, I see no reason to keep uids in share links.

$\endgroup$
  • 1
    $\begingroup$ re 2: from the opposite direction, this is what I meant by the current method is (fortunately) transparent. Your alternative method to me is a complete non-starter. Do you actually consider it as improvement if these references to content of the site were resolvable only using a private database and method owned by SE? It feels a bit ironic that concerns of surveilance (or even spying) by SE should be resolved by SE creating a new database of varied user-actions. The method that only UID is encrypted would be less problematic, although I'd also consider it as worse than the present. $\endgroup$ – user9072 Mar 13 '15 at 12:42
  • $\begingroup$ re 1: first, what you recall is not a plausible use-case, by contrast it is a (somewhat) plausible risk for a user-error (with negative consequences). The principal method to create reference to the site is via "cite," which provides also the link (and obviously without uid). Second, I for one, would not like a community were everything that is (mainly) for fun is forbidden, as soon as there is some risk of somebody by committing an error could somehow suffer some negative consequence. (In addition it is not only "for fun" but in some sense to help the growth of the site.) $\endgroup$ – user9072 Mar 13 '15 at 12:56
  • 1
    $\begingroup$ @quid I have replaced "use case" with the more neutral "scenario". Re 2: there are three possible goals here: (1) Privacy from SE: The SE owners cannot determine who clicked on a permalink (2) Privacy from everyone else: everyone else apart from SE cannot determine who clicked on a permalink (3) Gamification: SE can award Announcer badges. Clearly (1) and (3) are incompatible, so we have to decide which one we want. This is a political decision, and I am not taking sides. What I did is simply pointing out that (2) and (3) are compatible. $\endgroup$ – Federico Poloni Mar 13 '15 at 13:11
  • $\begingroup$ Sorry, I wrote the wrong sentence, please replace "who clicked on a permalink" with "who shared a clicked permalink". $\endgroup$ – Federico Poloni Mar 13 '15 at 13:17
  • $\begingroup$ Thanks for the clarification. A point of my reply was to point out some drawbacks of what you proposed, e.g., loss of transparency and additional dependance on SE. Now, for you more specific claims: at first it seems there is some merit to your claims of in/compatibilities but let us look at the situation in more detail as I find it interesting and possibly instructive. $\endgroup$ – user9072 Mar 13 '15 at 13:40
  • 1
    $\begingroup$ So, you propose we simply encrypt UserID or also the full link. But the annoucer badge is linked to the question that generated it, then revealing the identity of the sharer of the link or to be precise one sharer of a link to that question, but it should be rare there are competing ones. And if somebody really wanted to know who created some link, it would be as easy as just triggering a badge award (which is possible most of the time). $\endgroup$ – user9072 Mar 13 '15 at 13:47
  • $\begingroup$ This change creates the situation of complete pseudo-security. The situation is sufficiently non-transparent that looking locally one might actually believe one can anonymously share the link, while one really cannot. Now, you might reply, then let us scratch that list. But, first, somebody has to think of doing it; second, one looses functionality; third, if it is not anymore necessary to know for which question the badge was awarded then even 1 and 3 become somewhat compatible with sufficient effort (I think). But really, it seems ill-advised to me to go down that road. $\endgroup$ – user9072 Mar 13 '15 at 13:50
  • $\begingroup$ @quid re2: Why worry about dependence on SE? A referee choosing to include a link to SE in his/her report already depneds on SE servers' ability to fetch the right page with that URL, encrypted or not. A layer of crypto does not add much to that dependence. $\endgroup$ – Fan Zheng Mar 20 '15 at 20:19
  • 1
    $\begingroup$ @quid re1: I don't agree with you that failing to click on "cite" is an user error. As a newbie (which I believe many referees are), I don't even know where the "cite" link is. $\endgroup$ – Fan Zheng Mar 20 '15 at 20:21
  • $\begingroup$ @FanZheng: re 2: what you say is not quuite correct. The content is available also in data-dumps that can be (and are AFAIK) hosted on other servers too. Given a "clear" URL it is possible to retrieve the data from there as a back-up. The issue is with long-time support (not so relevant for a referee report but quite relevant for published papers). This is especially relevant for MO as there is a contract in place that would allow MO to leave the network with all the data, except for private/personal user-data. If MO should decide to do so, now links keep working, with crypto likely not. $\endgroup$ – user9072 Mar 20 '15 at 20:31
  • $\begingroup$ @FanZheng re 1: that somebody might not know it is an error, does not alter the fact that it is an error. Don't get me wrong, I will agree it is somewhat easy to make that error and likely something should be done to change this, but all this does not alter the fact that it is an error. (The "cite" link is in the pop-up once you click "share"; you will get a full bib-tex entry, including a link. On main only.) $\endgroup$ – user9072 Mar 20 '15 at 20:36
  • $\begingroup$ @quid re 2: I didn't quite see what you mean. So suppose MO is down, and you saw a link meta.mathoverflow.net/a/2181 in a paper. How do you retrieve the page independently of the SE servers? $\endgroup$ – Fan Zheng Mar 20 '15 at 20:38
  • 2
    $\begingroup$ @quid re 1: Don't you think a software that can easily lead people to make errors is, at the very least, crappy? Also I have explained in a comment below that this feature/bug/whatever is not easily available from FAQ/ help center/wherever you normally expect it to be. $\endgroup$ – Fan Zheng Mar 20 '15 at 20:46
  • $\begingroup$ @FanZheng For example go to archive.org/details/stackexchange, download the relevant archive, search for the post "2181" in there. Or, at some point somebody setup a read-only version of MO that was browsable just like the usual version and would make that convenient. This would be important in a case where for example SE would stop to exist (not to bridge a half-hour outage). It can be that for recent post there is no recent enough dump, but as said this is about longterm support. $\endgroup$ – user9072 Mar 20 '15 at 20:51
  • $\begingroup$ @FanZheng since you self-identified as newbie let me hereby inform you that you should normally expect to find documentation on the metas, especially on Meta Stack Exchange So what you could/should do if you want to know what "share" does is go there and search for it, like so meta.stackexchange.com/search?q=share then browsing the hits will inform you. The current top-hit by relevance essentially already informs you abut it. $\endgroup$ – user9072 Mar 20 '15 at 21:13
11
$\begingroup$

I consider the current situation as problematic, and would like to see this problem adressed by providing just a way to get direct-links without UserID (replacing the "share" options by a more simple link-feature).


This is a poll answer. Please do not downvote. Up-vote competing answers instead.

$\endgroup$
10
$\begingroup$

Guide to the poll answers

The general context is as explained in OP the fact that the link created and displayed when clicking "share" below a post contains the UserID of the user (when the action is performed by a logged in user).

The form of such a link is http://mathoverflow.net/x/PostNumber/UserID
where "x" is "q" or "a" depending on whether the post is a Question or an Answer (in actual reality "a" and "q" are interchangable). The link also works without the UserID part, and is created in this form if one is not logged-in.

Some general points to consider

  • Such links can lead to individuals revealing their identity and or MO account by accident, especially as the fact that they contain the UserID is not signaled when the link is created.

  • Links with UserID are used and more or less needed for some functionality of the site (albeit arguably rather tangential functionality).

  • The link with UserID seems like the default way to create a short link.

  • It is possible to get the essentially same link without UserID with one additional click ("share", then "cite.") [Added note: while the link alsways contains a "q" it will link to the answer, if it was created on an answer.]

  • The link can be turned into working links without UserID in a straightforward way by not including the number after the last slash.

The poll

The purpose of this poll is to get a feel on the community's opinion. The proposed options are:

Please upvote the particular answer(s) containing the option that matches your opinion. Please do not downvote poll-answers, to keep the vote-count transparent all the time for everybody.

If you find an important option missing please add it. However, please keep in mind that in any case this poll can only give a rough idea on opinions, and detailed ramifications and technical solutions would have to be discussed afterwards. So that additional answers similar in spirit yet different in the detailes might do more harm than good.

$\endgroup$
  • 1
    $\begingroup$ Two questions: Would it be simpler to make the poll in comments to one answer rather than in separate answers, especially since there are other answers as well? That would also make sure that there are no downvotes. If not, should the poll answers be CW? $\endgroup$ – Joonas Ilmavirta Mar 10 '15 at 17:41
  • $\begingroup$ @JoonasIlmavirta I thought about this but abandonned the idea as then editing is difficult/impossible and answers possibly added later might have even more disadvantage. They can be CW, but the sole effect is easier editing (which is possibly more relevant for the main post). Received votes are always irrelevant on meta, except for badges, which are then not touched by CW. $\endgroup$ – user9072 Mar 10 '15 at 17:45
  • $\begingroup$ One way to make a changeable poll in comments is to make short comments like "I'm in favour of option 1" and give the options themselves at the end of the question. I know that votes are somewhat irrelevant on meta, but I think CW status still has a psychological effect on voting. The "please do not downvote" request may seem more reasonable to many if it comes with CW. Anyway, the poll does seem reasonable to me, so don't take this as a complaint. $\endgroup$ – Joonas Ilmavirta Mar 10 '15 at 17:53
  • $\begingroup$ I made everything CW. I am really not sure about the comments. Perhaps it would have been better, but as the answers are posted already I think I do not want to change it now. $\endgroup$ – user9072 Mar 10 '15 at 17:57
  • $\begingroup$ I would slightly prefer the comment version (and there are not many votes yet), but it's ok as it is. I suppose sufficiently many people will figure out how the poll works even if the answers are mixed with this poll guide and other answers. $\endgroup$ – Joonas Ilmavirta Mar 10 '15 at 18:04
  • $\begingroup$ I added links in the hope to make this more clear. $\endgroup$ – user9072 Mar 10 '15 at 18:10
  • $\begingroup$ “It is possible to get the very same link without UserID with one additional click ("share", then "cite.")”: This is only true for questions, not answers. $\endgroup$ – Dmitri Pavlov Mar 10 '15 at 19:00
  • $\begingroup$ @DmitriPavlov what is your point: the link differs by having a "q" instead of an "a" I explain they are interchangable. Did you not understand this explanation or do you really want to insist on it not being the very same link but a link differeing by one character that has the exact same functionality. $\endgroup$ – user9072 Mar 10 '15 at 19:02
  • $\begingroup$ @quid: I'm afraid I don't understand your last comment. What is your proposed procedure for obtaining the link to an answer? $\endgroup$ – Dmitri Pavlov Mar 10 '15 at 19:05
  • $\begingroup$ @quid: Why did the second option split into two identical options (both 2 and 3 propose exactly the same changes)? $\endgroup$ – Dmitri Pavlov Mar 10 '15 at 19:09
  • $\begingroup$ @DmitriPavlov for the link, the procedure is the same for questions and answers. Please, can I ask you to just try to paste the link into your navigation bar before asking once again. $\endgroup$ – user9072 Mar 10 '15 at 19:11
  • $\begingroup$ @quid: I now see what you mean. Well, I guess it's nontrivial to realize this, especially because the title supplied by “cite” is the title of the question, and the link itself looks like a question link. $\endgroup$ – Dmitri Pavlov Mar 10 '15 at 19:14
  • 1
    $\begingroup$ @DmitriPavlov Good. I agree that this is not clear, and I already acknowledged twice (once yesterday and once today) that the situation for answers is not optimal when using "cite." (See in particular my most recent comment on OP.) This brings me to the splitting: one can find the current situation as not a problem while still seeing room for improvement and supporting such improvement. This is an opinion substantially different from the opinion that the current situation is a problem and there is need for improvement. $\endgroup$ – user9072 Mar 10 '15 at 19:21
  • 1
    $\begingroup$ @quid: Another question: is the phrase “Up-vote competing answers” (in plural) supposed to imply that users can upvote multiple poll answers instead of just one? $\endgroup$ – Dmitri Pavlov Mar 10 '15 at 20:01
  • 1
    $\begingroup$ @RicardoAndrade it is possible but then there are only two answers that are not part of the poll; and the poll answers themselves are really short. Moreover, I think not few users browse by active, and at the moment the poll-answers are nicely ordered on top. A possibility would be to have this "guide" accepted, for the momenet at least, so that it is on top. $\endgroup$ – user9072 Mar 12 '15 at 18:57
8
$\begingroup$

The behavior is documented.

Please see for example How do I link to a MO answer? to find this information (as a comment to the accepted answer).

Moreover, a moment's thought and the existence of the "announcer"-badge will show that some links most be non-anonymous.

It is easy to get an anonymous link via:

  • only copying a part of the link provided by share,

  • via not using share, but retrieving the URL in a different way (trivial for questions, and not hard for answers).

  • via not being logged in (as even mentioned in OP).

Thus, I consider this as a non-issue. Generally, when using a "share"-feature I would assume as a default that the user sharing is knowable. That in this case it is in fact the case, is made quite transparent; the URL is not obfuscated in any way.

$\endgroup$
  • 12
    $\begingroup$ You seem to be knocking down a straw man; I've never claimed that this behavior is secret or that it is hard to circumvent. The issue that you fail to address is that almost all users are not aware of this behavior. $\endgroup$ – Dmitri Pavlov Mar 8 '15 at 23:20
  • 41
    $\begingroup$ Vote this comment up if you were not aware of this behavior. $\endgroup$ – Todd Trimble Mar 8 '15 at 23:34
  • 26
    $\begingroup$ Vote this comment up if you were aware of this behavior. $\endgroup$ – Todd Trimble Mar 8 '15 at 23:34
  • 4
    $\begingroup$ @quid: A weaker statement will suffice: your claim about this problem being a nonissue only makes sense if all or almost all users are aware of this behavior, for which you supply no evidence. $\endgroup$ – Dmitri Pavlov Mar 8 '15 at 23:35
  • 7
    $\begingroup$ @ToddTrimble: An important factor to consider in all such polls is that visitors to Meta can be expected to be much more knowledgeable about MO than a typical user. $\endgroup$ – Dmitri Pavlov Mar 8 '15 at 23:39
  • 9
    $\begingroup$ quid, let's try to de-personalize this a bit ("first and foremost you that is to blame" addressed to Dmitri or to unwitting linkers seems pretty harsh). We don't have to get dramatic, but let's agree it makes sense to make this information much more visible. I don't believe Dmitri's scenario with the referee is all that contrived; in fact it's embarrassingly easy not to know these things, and it doesn't seem to be very visibly documented after all. $\endgroup$ – Todd Trimble Mar 9 '15 at 0:04
  • 7
    $\begingroup$ The harshness ("spying", "unethical" etc.) is, as far as I can see, directed primarily at a corporation, which I do not regard as a person. :-) I'm not sure why one would take offense particularly, but if we put that aside (or choose to dismiss it as overdramatic), then one is still left with a reasonable point, and that's what I think we should focus on. $\endgroup$ – Todd Trimble Mar 9 '15 at 0:13
  • 14
    $\begingroup$ I find the last sentence of this answer totally ridiculous; in what world is the meaning of the number 402 at a URL like mathoverflow.net/a/198099/402 "transparent"? $\endgroup$ – Eric Wofsey Mar 9 '15 at 8:42
  • 3
    $\begingroup$ @quid: I think there is a huge difference between a URL like mathoverflow.net/a/198099/402 and a (potential) URL like mathoverflow.net/a/198099/dmitri-pavlov. The latter URL makes it clear that the link is under surveillance and the former URL does not. The numerical user ID is not displayed “all the time” as you claim and few people know what their number is. $\endgroup$ – Dmitri Pavlov Mar 9 '15 at 12:46
  • 9
    $\begingroup$ @quid: You seem to disagree, but I think that a reasonable person copying a URL from the "share" button (whose mouseover text says merely "short permalink to this answer") might not even think of the fact that the link could somehow identify them if the URL does not appear to. I also don't think it's trivial to figure out what that number means if you don't recognize your user id (which, I imagine, many if not most MO users do not). $\endgroup$ – Eric Wofsey Mar 9 '15 at 12:47
  • 3
    $\begingroup$ @DmitriPavlov: As a minor technical point, your proposed URL wouldn't work, since names do not uniquely identify users (of course, this is trivial to fix). In any case, regardless of how transparent the URL is, I don't see a compelling motivation to identify the referrer by default at all. Yes, responsible users can notice and use an anonymous link, but why should we force them to do so at all? Is the data that SE gets from this really worth jeopardizing the privacy of careless users? $\endgroup$ – Eric Wofsey Mar 9 '15 at 13:14
  • 2
    $\begingroup$ @EricWofsey Your last question might be based on a false premise. In my understanding, the very point of "share" is to provide a way of generating a personalized link, so that users can get some credit (via the badge) for advertising the site. The intent is to encourage advertising the site by giving credit for doing so. I do not think SE cares specifically about data they get from it. What kind of data to what end would that be? I agree that the mouse-over text is not optimal. And feature-req on MSE to separate "link" and "share" $\endgroup$ – user9072 Mar 9 '15 at 17:19
  • 1
    $\begingroup$ @quid: I'd say that “share” simply implies some kind of posting to Twitter, Facebook, and other social networks. I'd not say that “share” implies that the resulting link will contain any personal information, apart from the information already contained in the identity of the poster on a social network. Perhaps we could organize another poll about this, but in my opinion the above poll already makes it clear enough that the majority does not think about “share” in the way you described. $\endgroup$ – Dmitri Pavlov Mar 9 '15 at 17:36
  • 5
    $\begingroup$ @quid: Fair enough, but I find it obnoxious that the only obvious way to get a link to an answer is to "share" it (yes, there are others, but it is substantially less obvious how to find them). Personally, I find the promotion of "sharing" as the default way to link to a MO post to be rather incongruous with our academic tone. But maybe I'm just a cranky old man who gets annoyed by social media. $\endgroup$ – Eric Wofsey Mar 9 '15 at 17:57
  • 2
    $\begingroup$ @quid so by documented you mean explained in another meta post? I guess only a handful of users ever went to that post to check it out before you gave that link, and only a handful of users will check out this post to know about the privacy leakage. I was expecting it to be explained in the FAQ. $\endgroup$ – Fan Zheng Mar 20 '15 at 20:33
7
$\begingroup$

Privacy concerns are legitimate, even though I do not feel that one has any factual basis on which to make claims of spying or unethical behavior, whether on the part of individuals or corporations. I'd rather not take the argument in that direction -- in fact I feel that would be fruitless unless more information comes to light -- but simply acknowledge that there should be greater transparency in this matter.

I'd also like to acknowledge quid's point that in fact the feature is documented publicly, and is there by design. I would also like to thank him (or her!) for linking to places where this is discussed. We can continue to discuss whether there is sufficient reason to include such numerical identifiers in short links, but I think we should also discuss for now possible remedies to the fact that a sizable number of people are (or were) unaware of the meaning of that number (which in my case is 2926), at least according to what my mini-poll under quid's post suggests.

It might be something so simple as having "what is this?" just after the numerical identifier, which one could click on and be taken to a short explanatory text, somewhat after the manner when one is asked to provide a CCV on the back of one's credit card while making an online purchase.

By the way, "Copy Link Location" is no different: a copy-and-paste also reveals the sharer's numerical ID. I hadn't noticed that before, even though I often link to an MO answer from the nLab using this utility. If that is a concern, then users should go to the address bar of the answer instead. I'll bet a lot of people don't know this.

$\endgroup$
  • $\begingroup$ I adjusted the OP to use the word “surveillance”, which seems to describe the situation more accurately (surveillance need not be secret). $\endgroup$ – Dmitri Pavlov Mar 9 '15 at 14:53
5
$\begingroup$

I consider the current situation as alright, but at the same time I see room for improvement; e.g., by making the nature of the link more transparent and/or providing direct-link without UserID in addition with the same visibility.


This is a poll answer. Please do not downvote. Up-vote competing answers instead.

$\endgroup$
3
$\begingroup$

An overwhelming majority of those who participated in the poll are in favor of modifying the existing system for obtaining links to answers. How does one proceed from here? Shall we put in a request to modify the SE software?

$\endgroup$
  • 1
    $\begingroup$ Probably next would be discussion on meta.stackexchange.com ... assuming there is no thread on it already. $\endgroup$ – Gerald Edgar Mar 1 '16 at 15:35
  • $\begingroup$ It seems you missed in any case you do not acknowledge something did happen in the time since the polls. Namely what is proposed in the clearly top-voted answer was done; "cite" now has same visibility as "share" and gives a direct non-personalized URL. Anyway, even if you should think the change that did happen is not sufficient I think you ought to acknowledge the change and argue why it is not sufficient. $\endgroup$ – user9072 Mar 1 '16 at 15:46
  • $\begingroup$ @quid: I fail to see how what was done has anything to do about the issue under discussion; recall that the top-voted answer talks about “making the nature of the link more transparent and/or providing direct-link without UserID in addition with the same visibility”. With respect to the first point, nothing was done. With respect to the second point, no direct link is provided in the citation, only a plain text URL that is not a hyperlink, again no changes relevant for the discussion. $\endgroup$ – Dmitri Pavlov Mar 1 '16 at 15:58
  • $\begingroup$ @GeraldEdgar: I don't see why such a discussion would be relevant; MO and MSE are very different communities with very different goals, audiences, etc. $\endgroup$ – Dmitri Pavlov Mar 1 '16 at 16:00
  • $\begingroup$ re 'only a plain text URL that is not a hyperlink' Well if this distinction now is important, then let us observe that there is no hyperlink under share but merely a plain text URL. Thus the entire thing seems to be moot from the very start on. ;-) $\endgroup$ – user9072 Mar 1 '16 at 16:12
  • $\begingroup$ @quid: Your claim is false: “share” is a hyperlink. $\endgroup$ – Dmitri Pavlov Mar 1 '16 at 16:29
  • $\begingroup$ I click "share" I get a window from which i can copy a plain text URL. I click "cite" I get a window from which I can copy a plain text URL. In that sense they are the same to me. Clicking it is the main use-case of "share" to me. Note that I said "under share." However I now see that there is some distinction in that "share" can be used more directly as link too (via right-click, copy-link); sorry for the confusion, but it never even occurred to me to try that. I am not sure about the technicalities; maybe the same functionality could be added for "cite" or this part of "share" disabled. $\endgroup$ – user9072 Mar 1 '16 at 17:00
  • 1
    $\begingroup$ On the general point, let me reiterate: in my mind there are now two options with basically equal visibility to get a direct URL one non-personalized; this addresses the .second part and as the main conjunction is "or" the thing is done. If you think otherwise, that's fine, but I think you still should acknowledge/address the change and the fact that something was done related to this. Then you can say what you want in addition since again in my mind the thing did get addressed. (Tangentially what do you mean by MSE? I suspect the math site, but this is not what G.E. meant.) $\endgroup$ – user9072 Mar 1 '16 at 17:03
  • $\begingroup$ @quid: Once again you're confusing links and URLs. My original question and the poll both talk about links, not URLs. There is no direct link under “cite”, so your claims about the second part being fulfilled are simply not true. Using an ultraverbalist interpretation of the particular formulation of a poll answer to claim that a minor tweak in the interface somehow addresses the deep underlying privacy issues, is, frankly, ridiculous. I'm also getting really tired of discussing the nuances of your formulations of poll questions; it's irrelevant and we're not in a law school here. $\endgroup$ – Dmitri Pavlov Mar 1 '16 at 17:29
  • $\begingroup$ I told you that what now appears to be the problematic "link" for you did never occur to me until an hour ago. Please read the start of the guide to the poll answer (with emphasis) "The general context is as explained in OP the fact that the link created and displayed when clicking "share" below a post contains the UserID of the user (when the action is performed by a logged in user)." Whether "link" is the right word, it seems clear to me what my intent was then, namely to talk about about what one gets when one clicks "share" and this is a window with a plain text URL to copy. $\endgroup$ – user9072 Mar 1 '16 at 18:06
  • $\begingroup$ This is the description based on which the answers where phrased and voting happened. If this did not match your intent then I am sorry but you'd better have raised this then. And, I'd be actually sorry if there truly was a misunderstanding. But again I was always talking about the content of the window, and this is what is written as context for the poll. $\endgroup$ – user9072 Mar 1 '16 at 18:10
  • $\begingroup$ Sorry for one more comment. I just browsed the older discussions again there are numerous instances (not just from me) where link and URL are used in a basically synoymous. “It is possible to get the very same link without UserID with one additional click ("share", then "cite.")”: This is only true for questions, not answers. – Dmitri Pavlov Mar 10 '15 at 19:00" You quote me yet you did not object to my use of link for plain text URL or find the distinction important. I think it was quite clear the main issue is getting an URL to copy. $\endgroup$ – user9072 Mar 1 '16 at 18:42
  • 3
    $\begingroup$ I assumed you wanted links of the form http://mathoverflow.net/a/198099/402 changed by default to omit the user number. I also assumed those links are used throughout the stackoverflow system, and such a change would have to come from them. Was I wrong? $\endgroup$ – Gerald Edgar Mar 1 '16 at 19:33
  • 1
    $\begingroup$ I rather doubt that most users use 'right-click on share, copy location' when they need the URL; for example note that nobody proposed this shortcut on meta.mathoverflow.net/questions/1885/… it also did not come up in this debate until very recently. But rather click and copy. Anyway, this is perhaps besides the point. I might agree that copying the URL from the share window is still slightly simpler than from "cite" but then that URL is given no less than three(!) times under "cite" it is really not that hard. $\endgroup$ – user9072 Mar 1 '16 at 20:05
  • 1
    $\begingroup$ All that said I thought that change was sufficient; I even had entertained the idea to bring it up here. If you think otherwise, fine, just post a specific feature-request, preferably in line with the topvoted answer (that is both types of URLs are preserved; the removal of this type of URL did not find much support) and maybe acknowledge the change that happened since this debate (or don't), and see what happens. $\endgroup$ – user9072 Mar 1 '16 at 20:08

You must log in to answer this question.

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