Last modified: 2011-10-21 23:03:14 UTC
It has came to my attention that a lot of people seem to be posting their email addresses in an text<image>text format. Bots are able to read emails from that. Instead perhaps an image should be generated with an extension displaying the text as an image. I was thinking of something like <math> but am not sure the best way to execute this. Perhaps the magic word would display whatever email user is confirmed with and post that provided user enables this in settings. This is intended as an anti-spam measure.
This would kill usability in a few ways... And there would be nothing stopping these bots from indexing/OCRing the images. If users don't want their address public and/or indexable, they can list the Special:EmailUser method ([[Special:EmailUser/<USERNAME>]]) on the page and users can contact them via that.
Kill usability how? People already use images as email... forexample the use of @ image. If people want to use the current method they would still be able to. This just provides a third method. If what you suggest (OCRing of images) happens, that would be a wonderful thing. It means the spam bots will be too busy OCRing images on userpages. This would waste their resources which is a positive thing for us.
(In reply to comment #2) > Kill usability how? People already use images as email... forexample the use of > @ image. If people want to use the current method they would still be able to. That template (or image) when done properly (en.wikipedia) already gives "@" as the alt text so when you copy and paste it still outputs the email address properly. If the whole address was a image, we would still want the alt text to be the full email address for users that don't display images (or using a screen reader package) so it would be no different than not using the image because the address would still be intact. Example of how they would be laid out: Without any image: joe@somebody.id.au With at symbol as image: joe<img src="..." alt="@" />somebody.id.au Full with alt: <img src="..." alt="joe@somebody.id.au" />
The idea here is to offer a third way for people who want to display their emails on their userpage but not want to be spammed like crazy. The idea is to confuse the bots. If people want to use the current method text<image of @>text.ext that would still be fine. Why do we want alt description? We could do without that particularly for emails. (and who doesn't display images really). If I do not want an alt description for my email, I should not be forced to display it. It could be crafted in a manner that the image becomes unreadable to bots. Image could be fragmented to make OCR impossible or very difficult. For instance OCR will have great difficulty recognizing a character if the character is broken in to two or more images. Or it could be one image per character. Also it is possible to add fake/invisible images in the mix that machines would pick up but not people. The more difficult it is for bots the better for us. I am fully aware there are ways around every precaution but that shouldn't mean we should surrender to spam bots.
(In reply to comment #4) > (and who doesn't display images really) - Blind users - People on slow connections who turn off images because it speeds things up - Likewise people on slow connections who defer to text browsers because they have less to load > It could be crafted in a manner that the image becomes unreadable to bots. > Image could be fragmented to make OCR impossible or very difficult. For > instance OCR will have great difficulty recognizing a character if the > character is broken in to two or more images. Or it could be one image per > character. Also it is possible to add fake/invisible images in the mix that > machines would pick up but not people. > > The more difficult it is for bots the better for us. I am fully aware there are > ways around every precaution but that shouldn't mean we should surrender to > spam bots. And real users should lose out on a valid bit of usability, the ability to not have to manually type in an e-mail address and instead click a link to send an e-mail. Because of spam? ...imho, that's like saying Wikipedia should turn anon editing off because of vandals.
People can still use wikipedias own email system mentioned if they desire to not deal with images or if they are blind. The email on userpage is just an additional means. The other option for people like me is to display nothing which is in essence less usable. Your example about ip edits does not correlate with this proposal, please stay on topic.
I don't think this is something we'd really need/want as a generalized feature in MediaWiki itself or on any Wikimedia sites. You can of course devise your own image that's as convenient or inconvenient as you like and upload it. If you're interested in writing an extension to help build it in for your own sites, we'd be more than happy to make sure it's hosted with our code repositories for extensions.
