The proposal seems mostly good. I like the idea of reusing Markdown headers and table syntax, rather than inventing yet another syntax.
As mentioned in the tables topic, what concerns me about Github Flavored Markdown tables is the column alignment feature. There are other ways you could achieve the same result without mixing presentation with content (and without using classes):
- Use nth-of-type in CSS to float the content of particular
<td>columns left or right. - The application could look at the data in the column and automatically align it left or right based on the data type that is found.
I don’t recall explicitly defining column alignments in CSV. Does Excel use the second approach?
Another question: if a PSV document contains content outside of the table (e.g. paragraphs and lists) would these be ignored or added to a cell on their own?