Babelmark shows that lists separated by blank lines are treated inconsistently. Changing the number of blank lines between lists leads to even more inconsistent results.
Test cases
These are two basic test cases:
One blank line (should produce only one list, if I interpreted the spec correctly):
* aaa
* bbb
* ccc
Two blank lines (should produce two lists):
* aaa
* bbb
* ccc
Expected results
I was expecting the first test case to produce just one list and the second to produce the following:
<ul>
<li>aaa</li>
<li>bbb</li>
</ul>
<ul>
<li>ccc</li>
</ul>
The issue
Almost no implementation produces the expected result and most implementations produce weird artifacts (e.g. extra <p> elements).
Babelmark2 for two blank lines
Some examples.
commonmark.js produces just one list, but with <li><p>aaa</p></li> items instead of the simpler <li>aaa</li> items that are produced when there are no blank lines.
pandoc produces a weird mix of <li></li> and <li><p></p></li>:
<ul>
<li>foo</li>
<li>
<p>bar</p>
</li>
<li>
<p>baz</p>
</li>
</ul>
marked produces two different results. With one blank line it produces exactly what pandoc produces, while with two blank lines it produces the correct (or expected)
<ul>
<li>foo</li>
<li>bar</li>
</ul>
<ul>
<li>baz</li>
</ul>