As well as agreeing to this, I would add the following:
* I would support both Python 2 and Python 3. Yes I know that the push to Python 3 is ongoing but there are a lot more libraries to be found for 2. Don't bother with IronPython.
* I would integrate pip or whatever package manager Python uses these days so that people can download their own modules and use them in the Excel sheets. Not only that, the packages should be restored automatically when the Excel sheet is loaded.
* It goes without saying that Python should be sandboxed.
* VS Code integration in Excel with full syntax highlighting and autocomplete - again a no-brainer.
Nah. They can spare a lot of pain and confusion by not having two different Pythons with different sets of libraries in Excel, and they have this opportunity because they don't have to support people's old Python 2 spreadsheets. There aren't old Python 2 spreadsheets.
The libraries that are stuck on Python 2 really don't seem relevant here. There aren't many of them anymore [1], and I don't think people will be trying to run graphite-web or some busted old OpenID provider in their spreadsheet.
(I originally used Fabric and Twisted as examples in the above, but hey, those are on Python 3 now. Also not things you'd be likely to use in a spreadsheet.)
Most of the big data science libraries are sunsetting support by 2020, so I don't think supporting Python 2 is a great idea. It used to be true that there were lots of libs stuck on Python 2, but I'm struggling to think of any major ones that'd be relevant to Excel that haven't moved forwards nowadays. No way it'd be worth the confusion of having two different versions of Python and the documentation headache that'd be. There's a case to be made for keeping Python 2 around for stuff like internal software, but you're not gonna be running that in Excel.
Not to get into a huge python debate. But this statement is just blatantly false.[0] Unless you are in a niche domain, the library you need either supports Python 3 or (increasingly, esp. among newer packages) is 3-only.
* I would support both Python 2 and Python 3. Yes I know that the push to Python 3 is ongoing but there are a lot more libraries to be found for 2. Don't bother with IronPython.
* I would integrate pip or whatever package manager Python uses these days so that people can download their own modules and use them in the Excel sheets. Not only that, the packages should be restored automatically when the Excel sheet is loaded.
* It goes without saying that Python should be sandboxed.
* VS Code integration in Excel with full syntax highlighting and autocomplete - again a no-brainer.