8

Bug 1: Association bonuses for no apparent reason.

This user joined our stack with 101 rep (from the association bonus). On checking rep I see one other stack with 101 rep, and one stack with 1 rep.

The 101 on the other stack seemed a bit coincidental (exactly 10 upvotes? Possibly.) so I checked -- no activity. So that was an association bonus too.

On the third site, the user's page is 404.

So then I thought "well, maybe there's a temporary ban and 404/1 rep is an artifact of the ban". So I searched that site several ways; nothing.

Bug 2: site accounts separated/disconnected from each other

What I did find was a fourth site account that is "islanded" from the others, it doesn't list the other 3 stacks and they don't list it. (note the user icon is the same; purple chevrons). Again the association bonus; again not enough site activity to support it.


As tinkeringbell discusses, it's proving the "islanded" site account is actually a hidden account. I discovered this through a general web search on a [well-behaved][4] search engine. So this may be morphing into a bug report/feature request to make that more difficult.

0
11

People get to keep their association bonus, because that isn't tied to the profile you gain 200 reputation on but to the network account:

Even if you delete every single profile and the one on stackexchange.com, your underlying network account still exists, and that is where the association bonus is stored.

So, in this case there's a deleted profile that once had over 200 reputation, which now makes sure that anytime this user creates a profile for a site, they earn the association bonus.

The datascience profile looks deleted, although it's kinda weird it's still showing up. There does seem to be some kind of caching involved that may mean this still needs to be deleted from the list of network profiles.

As for the second 'bug', that's caused by hiding communities. Moderators can still see these, that's how I know for sure.

4
  • 1
    So moderator on one site can see all hidden communities on all sites? Or just the site they moderate? (e.g. if you're a mod on IPS, can you see that a user on Cooking has Gaming hidden?) Aug 17 '20 at 12:25
  • @Shadow if we know how to get to a network profile, we can see which ones are hidden, regardless of which site we mod. So yeah, I can see a user on Cooking has Gaming hidden, but I'd have to jump through the same hoops as regular users to get to that network profile for your hypothetical Cooking user. There's quite a few big hoops described out there ;)
    – Tinkeringbell Mod
    Aug 17 '20 at 12:33
  • For example, for the hidden 3d printing profile Jon describes there, I'd have to use something like what's described here to get to the network profile and see all hidden communities.
    – Tinkeringbell Mod
    Aug 17 '20 at 12:35
  • 2
    The association bonus was earned about a year ago on The Workplace.
    – Catija StaffMod
    Aug 17 '20 at 16:21
5

As Tinkeringbell discusses, I am seeing things meant to be hidden.

I solve problems like that for a living, so I can speak to how that happened and what to do about it. I discovered this in a general web search engine that is well-behaved, meaning it obeys the rules for directing both web spidering and content inclusion in search engines. Note these are voluntary rules and a blackhat spider may not obey.

How the spider found it

  • it found a contemporary link to the page somewhere on StackExchange,
  • it found a contemporary link on a third-party web site,
  • it already knew about the page's existence from past crawling.

I broke those out into 3 cases so I can address common misconceptions.

The second case is one where a webmaster (i.e. StackExchange) doesn't create any internal links, but an outsider does. Suppose someone's user page links to their Web page at sfu.edu. Unbeknownst to them, SFU publishes aggregate data from their web logs, including Referer. The search engine considers these to be a rich data source, for URL discovery (although not for link relevance aka PageRank). SFU uses rel=nofollow, but that is not a "do not follow the link" directive, that is a "disregard link relevance" directive.

The third case is the likeliest one here. Once a search engine has a URL, it doesn't drop the URL when all links to the URL disappear. That does have an effect on link relevance, but that only affects ranking. If the searcher is specific enough to where only dozens of results exist, even a low-ranking page will surface.

How to hide it from well-behaved engines

... Assuming StackExchange wishes to. There are arguments both ways.

The best option is the "noindex" HTML meta-tag (or X-Robots equivalent in HTTP), as such:

  <meta name="robots" content="noindex">

This tag says "do not include this page in search results". The tag can also have elements such as nofollow (which actually does mean "the crawler should disregard links found on this page", unlike rel=).

Note that the search engine has to fetch the page in order to see this tag; that's nothing to worry about. I suppose you could serve it in the X-Robots HTTP field and serve blank page content. However at that point you could just as easily return 403 Forbidden, 404 Not Found, or 410 Gone. Note that a 410 Gone means the URL is planned to be gone forever; this matters to some search engines.

Blocking the engine from crawling a page, e.g. excluding the page in robots.txt, will not work. Robots.txt only directs crawling, not indexing. A search engine may still create a search result for the page, inferring the page content from links and archived data.

You must log in to answer this question.

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