HackerNews links don't work

[edent]

Tried to submit https://news.ycombinator.com/item?id=35927509 to https://shkspr.mobi/blog/2023/05/the-new-zip-tld-is-going-to-cause-some-problems/

I got the error

Not enough data available

[dshanske]

You are right, but there isn't enough data. We may have to figure out how to find some.

[edent]

Same issue on Lemmy - for example https://lemmy.one/post/78196

[pfefferle]

This is the same issue as with hackernews, lemmy does not really use WebSemantics: http://php.microformats.io/?url=https%3A%2F%2Flemmy.one%2Fpost%2F78196

[dshanske]

Anyone want to read the HTML and see if there is something we can extract of any type?

[pfefferle]

@edent do lemmy and hackernews support Webmentions or have you simply tried to send them manually?

[edent]

@pfefferle I was submitting them manually to my site. That's where I saw the error.

Looking through the HTML of both HN & Lenny, I can see the links to my site. Is there something specific I should be looking for?

[pfefferle]

There is a bug in 5.0.0 that does not parse the meta-headers correctly. Maybe we wait until 5.1.0 is released to re-check.

[dshanske]

It's still a problem of what data could we extract to render a preview.

[edent]

So, if I've understood correctly, the problem is that

wordpress-webmention/includes/Entity/class-item.php

Line 352 in 6e4ed42

public function verify() {

Either can't find an author or can't find a summary.

Changing verify() so the author always returns true gets me this blank comment:

For a basic WebMention, is it necessary to have a summary? What I'd like to appear on my site is:

This article was mentioned on [WebSite Name](http://example.com)

Could the name could come from the <title> or perhaps the domain name? I think that would be enough.

[janboddez]

Sorry to chime in like this. :-S That's what IndieBlocks does, too. Set a default name (either "Anonymous" or the page's host name) and content. Then if there are microformats, we turn those into something more meaningful.

Shouldn't be too hard to do something similar here. I'd be happy to look into it, and maybe submit a PR.

What I do for" default" mentions (and replies, etc.) is, where possible, actually use the code that WP uses for pingbacks, set the comment to something like "[...] some text linking back to this article [...]" and otherwise fall back to "mentioned this" or something.

Not saying Webmention should do the same, but that could perhaps be an option, too.

[dshanske]

I think we can add something simple.. just was trying to avoid Anonymous as it looks like less useful info. I think Hackernews may be worth a custom rule to find things.

[pfefferle]

Maybe use the domain host?

[edent]

Also, LinkedIn (urgh!). Take a URl like this - https://www.linkedin.com/posts/terenceeden_why-im-using-mx-as-a-title-activity-7074702026084884480-jO0C?utm_source=share&utm_medium=member_desktop

It doesn't contain the direct link in the HTML, but it is in the JSON+LD as "sharedContent":

<script type="application/ld+json">
{
  "@context": "http://schema.org",
  "@type": "SocialMediaPosting",
  "@id": "https://www.linkedin.com/posts/terenceeden_why-im-using-mx-as-a-title-activity-7074702026084884480-jO0C",
  "author": {
    "name": "Terence Eden",
    "image": {
      "url": "https://media.licdn.com/dms/image/C4E03AQEX11qBnAo43A/profile-displayphoto-shrink_400_400/0/1517677542848?e=1692230400&v=beta&t=-AHOoj__Ehm_p24dQ6McPbqNj9gQ2UsjvzOCIU16IGs",
      "@type": "ImageObject"
    },
    "url": "https://uk.linkedin.com/in/terenceeden",
    "@type": "Person"
  },
  "datePublished": "2023-06-14T11:00:04.600Z",
  "image": {
    "url": "https://media.licdn.com/dms/image/sync/D5627AQHWF4nuGqlfzQ/articleshare-shrink_800/0/1686397225711?e=1687428000&v=beta&t=SsQ_1b8M3xAGXnhUvID0976rpItlTLdC2Gz9RG3sXwI",
    "@type": "ImageObject"
  },
  "sharedContent": {
    "@type": "WebPage",
    "headline": "Why I'm using \"Mx\" as a title",
    "url": "https://shkspr.mobi/blog/2023/06/why-im-using-mx-as-a-title/"
  },
  "articleBody": "What title do you use when filling in forms? Mr? Mrs? Ms? Or something newer and more incluside?\n\nWhy do companies want to know my title? What are they going to do with my data?\n\nHere's why I'm using \"Mx\" as my title.\nhttps://lnkd.in/eCNH5sqp",
  "isAccessibleForFree": false,
  "hasPart": {
    "@type": "WebPageElement",
    "isAccessibleForFree": false,
    "cssSelector": ".details"
  }
}
</script>

[pfefferle]

The JSON-LD parser had a bug, but should work with the latest PR.

[edent]

Cool! When there's a new release I'll give it a go and report back.

Thanks for all your work on this.

[edent]

@pfefferle LinkedIn is now working - thanks :-)

[pfefferle]

Yea, hacker news is a bit tricky, because there are nearly no informations we can use.

[dshanske]

There's always the API. https://github.com/HackerNews/API

[pfefferle]

But I wouldn't see that as part of the plugin! that's too specific! 😳

Any ideas what we could add to make Hackernews at least pingable and not looking too ugly?

[edent]

Grabbing the text from the <title> element seems like the easiest solution to me. Or use the domain name as a fallback?

[pfefferle]

Result is now

{
  "published": {
    "date": "2023-12-28 19:41:45.864580",
    "timezone_type": 3,
    "timezone": "UTC"
  },
  "url": "https://news.ycombinator.com/item?id=35927509",
  "author": {
    "type": "card",
    "name": "news.ycombinator.com"
  },
  "name": "The new .zip TLD is going to cause some problems | Hacker News",
  "site_name": "news.ycombinator.com",
  "content": "The new .zip TLD is going to cause some problems | Hacker News",
  "response_type": "mention",
  "raw": {
    "referrer": "origin",
    "viewport": "width=device-width, initial-scale=1.0"
  }
}