Support HTML tags in markdown

[Aloso]

Is your proposal related to a problem?

CommonMark markdown supports HTML tags. Websites usually have an allowlist of accepted tags (here's the list of tags accepted by GitHub). Since Lemmy doesn't follow the CommonMark spec, it doesn't support HTML, but I think it should, since HTML is much more flexible and powerful.

Describe the solution you'd like

Lemmy should allow HTML tags. Ideally, it should also follow the CommonMark spec as much as possible.

Describe alternatives you've considered

None.

Additional context

Official CommonMark website and playground.

[dessalines]

Our markdown parser supports html but we don't have it turned on. To me it seems pretty dangerous. Do you have a use-case that goes beyond what markdown can do?

[Aloso]

It's not dangerous if only some html tags and attributes are allowed.

Do you have a use-case that goes beyond what markdown can do?

Yes. The most common use case is a table with line breaks or horizontal lines:

This is	a table with a line break.

HTML also allows you to create a table without a header:

A table

without a header

Table cells spanning multiple columns or rows:

Col 1	Col 2	Col 3	Col 4
3 columns			2 rows
2 columns		single

Table captions:

Monthly savings

Month	Savings
January	$100

Image width set to 25%:

Image captions:

Fig. 1: The Meta logo

Description lists:

Coffee

Black hot drink

Milk

White cold drink

Hover titles:

Hover me

Alignment:

I'm far-right!

I'm in the middle!

[Aloso]

Here's the code github uses for sanitizing HTML:

https://github.com/gjtorikian/html-pipeline/blob/main/lib/html/pipeline/sanitization_filter.rb#L44-L106

[dessalines]

We wouldn't have to do this, it's done by our front end markdown parser: https://github.com/markdown-it/markdown-it

Its fairly sure its safe to turn it on, its a good library.

[Nutomic]

It would also have to be enabled on the backend for RSS and federation.

[dessalines]

I don't know that anything would be required for comrak. It doesn't actually render that html and would probably leave it alone when running markdown_to_html.