Well that's interesting. It works in the notebook but not in nbviewer.

Yup - ANSI escapes are handled in javascript, which would need to be reimplemented in Python for nbconvert (or more javascript added to the nbconvert output). This is not really an nbviewer issue so much as an nbconvert one.

Ah, ok. That makes sense.

A lot of these issues seem to be nbconvert issues.

On a side note, it seems like nbviewer is always going to be the gateway to issues for nbconvert since this is the ipy in the sky.

Perhaps the error page should give people a simple way to run nbconvert on the notebook and write an issue within https://github.com/ipython/ipython.

500 errors are always nbviewer issues - that shouldn't happen. 'Wrong' rendering is typically nbconvert, and thus belongs at ipython proper. I don't think there's a problem pointing here by default, and then forwarding Issues to IPython when we know more about the culprit.

I don't think there's a problem pointing here by default, and then forwarding Issues to IPython when we know more about the culprit.

I agree...

Updated ANSI Test.ipynb location for reference: http://nbviewer.jupyter.org/github/jupyter/notebook/blob/master/tools/tests/ANSI Test.ipynb

I believe the underline and bold examples are the only ones not working on nbviewer today based on my quick side-by-side comparison of that nbviewer link versus the same ipynb rendered by a local notebook server.