Notebooks become 'not trusted' when code cells are modified in the paired file

[mwouts]

Consider the following notebook made of two cells:

import plotly.offline as offline
import plotly.graph_objs as go
offline.init_notebook_mode()

data = [go.Bar(y=[1, 3, 2])]
offline.iplot(data, show_link=False)

The notebook may become 'not trusted' when we modify the associated text representation (e.g. the associated py:light script).

Precisely,

• If change the cell content in a way that preserves the output, e.g. if I remove the space between 3, and 2, then the plot is preserved and the notebook remains trusted
• If I add new markdown cells, the notebook remains trusted
• But when I add a new code cell in the py:light script, or if I modify the plot other than by adding spaces, then the notebook becomes not trusted.

[mwouts]

@stefanuddenberg, do you think the above explains the issue you reported at #12?

[mwouts]

The above issue would be solved if we decide to trust notebook for which outputs are either empty (updated or new code cell), or trusted (code cell with unchanged output). This seems to be consistent with the jupyter security model.

[stefanuddenberg]

@stefanuddenberg, do you think the above explains the issue you reported at #12?

My problem is a little different; the notebooks are becoming un-trusted even when no changes have been made to any representation of the notebooks at all. But fixing this issue sounds like it might end up fixing mine in the process.

[mwouts]

I see. I think we should specifically address your issue then. Do you happen to have this issue with every notebook? Could you please test a notebook with just one cell with javascript output, say,

import plotly.offline as offline
offline.init_notebook_mode()

Thanks!