Make the types of HTML tag checked selectable and document that all are checked by default

[mmuehlfeldRH]

Summary

If a web site embeds an SVG image by using <embed type="image/svg+xml" src="file.svg">, linkchecker incorrectly interprets the file in "src" as a link.

Steps to reproduce

1. linkchecker --recursion-level=1 https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/configuring_and_managing_networking/getting-started-with-tipc_configuring-and-managing-networking

Actual result

URL        `images/TIPC-architectural-overview.svg'
Parent URL https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/configuring_and_managing_networking/getting-started-with-tipc_configuring-and-managing-networking, line 4989, col 299
Real URL   https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/configuring_and_managing_networking/images/TIPC-architectural-overview.svg
Check time 2.564 seconds
Size       9KB
Result     Error: 404 Not Found

Expected result

Do not interpret <embed type="image/svg+xml" src="images/TIPC-architectural-overview.svg"> as a link.

Source code snippet of the example page mentioned above

<div class="informalfigure">
  <div class="mediaobject">
    <object type="image/svg+xml" class="svg-img" data="https://access.redhat.com/webassets/avalon/d/Red_Hat_Enterprise_Linux-8-Configuring_and_managing_networking-en-US/images/ddad5eeb9b394bb21e78c9d605c4842b/TIPC-architectural-overview.svg">
      <embed type="image/svg+xml" src="images/TIPC-architectural-overview.svg"><!--Empty--></embed>
    </object>
  </div>
</div>

Environment

• Operating system: Fedora 38 x86_64
• Linkchecker version: 10.2.1
• Python version: 3.11.4
• Install method: RPM from Fedora repository

[cjmayo]

In what way is this incorrect?

All elements that can contain links are checked:

linkchecker/linkcheck/htmlutil/linkparse.py

Lines 29 to 39 in f744604

	LinkTags = {
	'a': ['href'],
	'applet': ['archive', 'src'],
	'area': ['href'],
	'audio': ['src'], # HTML5
	'bgsound': ['src'],
	'blockquote': ['cite'],
	'body': ['background'],
	'button': ['formaction'], # HTML5
	'del': ['cite'],
	'embed': ['pluginspage', 'src'],

[mmuehlfeldRH]

For me, as a normal web site user, a link is what I can click on an HTML page and that brings me to to a different page or an anchor on the same page (mostly <a href...>). I didn't expect that the tool interprets "link" as "every absolute and relative path in each tag that supports such paths", and it's also not defined in the docs.

How about adding the list with the tags to the man page to clarify what is tested:

CHECKED TAGS

Linkchecker verifies links in a wide range of HTML tags:

Link tags and attributes:
   a: href
   applet': archive, src
   area: href
   audio: src
   ...

Anchor tags and attributes
...

WML tags and attributues
...

Since this list very likely doesn't change frequently, it might be OK to simply copy/paste it to the docs.

What would also be cool: If the user could pass an option with the list of tags to check to linkchecker. For example: -l a,button,del. Of course, this option would only accept tags that are in the list in the source code with their attributes. Then users could decide on their own, and it would significantly speed up testing if a user only cares about links in a few tags.

[cjmayo]

link does actually mean the same to me. Just 20+ years of history to look after! And I don't think URLchecker would have been such a great name.

I guess the intent was that every attribute that can take a URL is checked.

Let's keep this ticket as a feature request.