Is it possible to “unlink” a link via something like an escape-sequence?
Cause: for example the university graduations M.Sc and B.Sc (master/bachelor of science) are automatically linked. I suspect “.sc” is read as the top-level-domain of the Seychelles.
I don’t think that this is a bug: presumably it’s not possible to separate wanted from unwanted links technical in a perfect way.
In 1993, the Internet Assigned Numbers Authority (IANA) explicitly reserved all single-letter and single-digit second-level domain names in the top-level domains com, net, and org, and grandfathered those that had already been assigned
The complete list of grandfathered domains from 1993 is i.net, q.com, q.net, x.com, x.org, and z.com
So it is true on the “traditional” TLDs that single letter domains effectively don’t exist outside the 1993 grandfathered list of 6. But it sounds like on the country TLDs it is more like the wild west…
The list of Country code top-level domains which have been identified to allow single-letter domains are : .ac .af .ag .ai .am .asia .biz .bo .by .bz .cm .cn .co .com.ar .com.br .co.at .co.il .co.uk .cr .cz .cx .dj .de .dk .fm .gd .gg .gl .gp .gs .gt .gy .hn .ht .ie. im .io .je .kg .ki .kw .la .lb .lc .ly .md .mg .mk .mp .ms .mW .mx .mu .nF .nz .pe .ph .pk .pl .pn .pr .pw .ro .sh .st .tc .tl .tt .to .tv .travel .ua .ws .vc .vg .vn and .vu
I have a deeper objection here, which is that we shouldn’t be auto-linkifying every single one of the zillions of TLDs, no matter how rare they are @tgxworld – that is much more risky.
I think if you want crazy-town TLD to auto-link, you should learn how to use the link button in the editor, or the markdown link codes.
This is nothing to do with @tgxworld here, this is linkify by markdown it, we have a few edge cases that pop up but nothing compared to previous linkify
A no-win situation. Passing text through a “known atypical domains” matcher feels like it would be unnecessarily costly. Considering that there are ways to linkify text in the UI that relative non-techies should be able to discover and delinkifying non-URLs is unexpected and requires some techiness to fix, I lean towards the not linkifying atypical URL side
My personal opinion, fuzzy links should be disabled.
But if you can’t live without fuzzy links - don’t extend TLD-s list to “everything valid”. If you allow all valid TLDs in fuzzy links, you will get a lot of false positives.
I am going to put a pr-welcome here for the community on
A site setting for “markdown_auto_link_tlds”
Sane defaults for English and other locales we support (per local defaults)
linkify.it should not be a link in English out-of-the-box, but it should probably be a link in if you are running in Italian out-of-the-box.
For default install in English I would autolink:
test.edu, test.gov, test.com, test.com.au, test.co, test.co.uk, test.ca. I am not sure if I would do test.sg, test.ph and test.in out-of-the-box despite the English speaking population. I am also unsure about enabling the fancy domains people use like bit.ly and bla.community it just feels like its not to onorous to add the https:// in front.
x.sc and so on should definitely not be on by default for any of our locales even French cause it is high risk of mistakes
Do people really type domains + paths manually? IMHO that’s the only case when fuzzy links could be useful.
If one copy-paste browser url, it’s detected by http(s) prefix and you only need to beautify it (drop protocol, maybe drop www prefix, drop the middle part for very long links)
Misplaced autolinking is a common issue in our community, where people use their smart devices to pick out posts and often leave out spaces after periods. This leads to alot of strange, errant autolinked text that requires editing later by moderators.
We would certainly turn this off if it became an admin setting.