A user, jtsagata, reported a problem with oneboxing links containing non-Latin characters in the URL. The issue prevents the links from being clickable in the onebox version, but not in the inline version. Falco suggested that the URLs are not valid, being Internationalized Resource Identifiers (IRIs) instead. jtsagata explained that these types of URLs are commonly used by local sites, including the Greek Wikipedia, and requested a solution to either pass the link as given or disable oneboxing for such cases.
Codinghorror provided two possible solutions: adding a space before the link or blacklisting the domain from oneboxing in the site settings. Lionel-rowe further analyzed the issue, explaining that the problem arises from double-encoding the URL.
Falco suspected that this might be a regression of a previously reported issue, and Lionel-rowe confirmed this, providing an example of a similar issue with a Russian Wikipedia link. Sam assigned the task to tgxworld for further investigation.
Finally, jtsagata reported that the problem was solved in the latest version.
Onebox have problem with some links from a wordpress site. The problem seems to have something with not latin characters in the url.
2 links to demonstrate the problem is this and this. I can follow the links in the inline version, but not from the onebox version. Same if i use a url shorten service.
The same links onebox version
Those aren’t valid URLs.
They are Internationalized Resource Identifier - Wikipedia which don’t get the same treatment.
Valid or invalid many of the local sites use this form and that is outside our control. For example all of the Greek wikipedia
And as everything seems to work, and the onebox is created it was not detectable by the end user. Why not just pass the link as given, or the short url as given;
Is there any way to detect and don’t make onebox it such cases, or to disable onebox completly?
Yes, trivially, just enter a space (or any other character or text) before the link.
You can also blacklist the domain from oneboxing in your site settings.
It doesn’t look like it. Raw:
https://cerebrux.net/2019/05/28/brave-browser-%ce%bf-%ce%b4%ce%af%ce%ba%ce%b1%ce%b9%ce%bf%cf%82-browser-%cf%80%ce%bf%cf%85-%cf%80%cf%81%ce%bf%cf%83%cf%86%ce%ad%cf%81%ce%b5%ce%b9-%cf%84%ce%b1%cf%87%cf%8d%cf%84%ce%b7%cf%84%ce%b1/
In the onebox link, this gets double-encoded to this:
https://cerebrux.net/2019/05/28/brave-browser-%25ce%25bf-%25ce%25b4%25ce%25af%25ce%25ba%25ce%25b1%25ce%25b9%25ce%25bf%25cf%2582-browser-%25cf%2580%25ce%25bf%25cf%2585-%25cf%2580%25cf%2581%25ce%25bf%25cf%2583%25cf%2586%25ce%25ad%25cf%2581%25ce%25b5%25ce%25b9-%25cf%2584%25ce%25b1%25cf%2587%25cf%258d%25cf%2584%25ce%25b7%25cf%2584%25ce%25b1/
The confusion may stem from the fact that Chrome and Firefox both display URIs as IRIs, e.g. in the address bar, for ease of reading.
Oh I see now, so it’s a report of a regression of Onebox breaks if there's chinese text in URL - #13 by tgxworld ?
I’m confused, because the oneboxes in OP are working fine for me in both Firefox and Chrome…
Looks like it. I guess the cooked versions are cached in that thread, but here’s the same link:
The onebox looks fine, but upon clicking the link, I get an error page (“Недопустимое название”) in both Chromium 74 and Firefox 67 (Ubuntu 18.04.2).
@tgxworld adding for your list to look at this, not a high priority but since you worked on the original makes sense you take this one.
Thanks for solving the problem in the latest version