[DISCUSSION] Consultation about effort required to fork and support this project

[sassdawe]

Greetings @claudiaregio,

I understand the deprecation decision was based on telemetry and priorities, etc. I can accept that even when personally I am not happy about it.

As you wrote in #4165:

Polyglot Notebooks and .NET Interactive are open‑source projects. The community is free to fork and continue them at any time.

Let's say the community would like to consider this option. May I ask the team to help the community better understand what went into publishing the dotnet global tool and the extension?

I have been looking into the project, reading the docs, etc., and I understand that there is a significant amount of pluming that is implemented outside of GitHub. Could the team share details about those things with the community - or if not with the community - could you share those things with MVPs under the NDA of the MVP program?

I ask because I'd like to keep using this product, and before any decision is made - more information would be required about what would it take to fork and ship a community version.

Thank you for the consideration.

Best regards,
David Sass
Microsoft PowerShell MVP

[jonsequitur]

I would be happy to chat about this.

[lextm]

Another open source project (Verso) is attempting to take things over,

https://www.reddit.com/r/dotnet/comments/1rmlwcr/i_built_verso_an_open_source_interactive_notebook/

[BearCarpenter]

I don't understand you, Microsoft. Why are you consistently abandoning F#? It's currently a great alternative to Rust. Those who don't understand borrow checker are looking for alternatives. F# is a very good one, the only one missing is DX(tools)!

Gleam, Moonbit, Zig, Rust, Pydantic, Typescript etc - look - we need language for rapid prototyping with types, monads, match, and efficient -> F#

[lextm]

@BearCarpenter F# is important, but far from important enough to influence the trends (many thousands on F# vs. millions on C#).

[sassdawe]

I would be happy to chat about this.

Hey @jonsequitur, thanks for the offer. In your opinion what would be the most appropriate channel for this to happen?

Maybe start from the assumption that we might touch something which is under NDA and clean/publish the transcript or notes after it?

Thanks!

[TorreyBetts]

Another open source project (Verso) is attempting to take things over,

https://www.reddit.com/r/dotnet/comments/1rmlwcr/i_built_verso_an_open_source_interactive_notebook/

Thanks for the mention. It's impossible to fill the shoes that the .NET Interactive team created, their work powered our learning in universities, careers, and personal life. It's something a lot of us have grown to love and grown accustomed to. I created Verso to help fill the gap and to take their original ideas further.

The founding principle is that every built in feature was made using the same public extension interfaces available to everyone. If a built in feature can't be built on the public API, then the API is considered incomplete.

VS Code Extension: https://marketplace.visualstudio.com/items?itemName=Datafication.verso-notebook
GitHub: https://github.com/DataficationSDK/Verso

[sassdawe]

I created Verso to help fill the gap and to take their original ideas further.

Hey @TorreyBetts, thank you for Verso Notebook! It looks interesting and I think we all appreciate your efforts which went into making it a standalone project and making it available us to use in such a short period of time.

However I feel like I need to put on my work hat and point to #4176 because I have fallen into the trap that I thought Polyglot Notebook had a similar support model like dotnet and PowerShell to make it usable at places where a component can't be abandoned with only 6 weeks of notice.

Similarly, I believe many companies place a higher level trust towards Microsoft made / maintained software than those created by other companies or the open-source community, see #4174. Still wearing my work hats: if I can't use Python because IT Risk Management considers it too risky... I am not sure how can I advocate to allow the usage of Verso.

Considering these aspects, I have been looking into what it would take to keep at least the level of SDLC which is already in place for dotnet interactive while figuring out how could we at least keep it on life-support while providing as much transparency related to its security as possible.

I also think the replacement should come with support of or with backing by something bigger, for example maybe by the .NET Foundation itself - because I'd prefer to only have to replace Polyglot Notebook once and not every 2-3 years.

[BearCarpenter]

@BearCarpenter F# is important, but far from important enough to influence the trends (many thousands on F# vs. millions on C#).

@lextm Based on the loss of an interesting tool in the ecosystem, one could draw the opposite conclusion – f# is becoming less and less important.

[TorreyBetts]

Similarly, I believe many companies place a higher level trust towards Microsoft made / maintained software than those created by other companies or the open-source community, see #4174. Still wearing my work hats: if I can't use Python because IT Risk Management considers it too risky... I am not sure how can I advocate to allow the usage of Verso.

That's a fair concern, and I appreciate you raising it. Organizational trust and longevity matter. A few things worth noting for anyone in the position of these IT risk management decisions: Verso is MIT licensed and fully open source, so the code is auditable and forkable. It has zero telemetry or phone home behavior, which is actually a simpler security story than many Microsoft tools. Every extension can be individually enabled, disabled, or even removed, so IT can control exactly what runs. Verso's dependencies are almost entirely Microsoft and .NET Foundation packages (Roslyn, FSharp.Compiler.Service, PowerShell SDK, ASP.NET Core). The full list is in our THIRD-PARTY-NOTICES.md. Supply chain risk isn't zero, but the dependency surface is small and auditable.

To finish, here's an example, say an IT department does not want to allow the running of Python for whatever reason. For Verso, that's as simple as removing the referenced assembly/project. The headless engine doesn't care that it's missing, or even knows about it. The engine is the orchestrator loading the available extensions at startup, which are the ones referenced by the thin client wrapping the Verso engine.

[sassdawe]

Verso is MIT licensed and fully open source, so the code is auditable and forkable.

I mean, this was the promise of dotnet interactive as well 🙃

From supply chain management standpoint nothing would improve for me - I am still wearing my "work hat" here btw - as forking and modifying Verso - or dotnet interactive for that matter, in order to make any of them an internal solution will require certain decisions and commitments from every company who might need a notebook experience.

Let me ask this: From your reddit post I know that Verso was part of a bigger SDK before and you had to pull it out of it, but if you didn't have that SDK to begin with, would you have started to write Verso from the ground up when the deprecation was announced or would you have forked dotnet interactive instead?

[TorreyBetts]

Let me ask this: From your reddit post I know that Verso was part of a bigger SDK before and you had to pull it out of it, but if you didn't have that SDK to begin with, would you have started to write Verso from the ground up when the deprecation was announced or would you have forked dotnet interactive instead?

Polyglot Notebooks started as a Jupyter Kernel. Now it's currently in a hybrid state where the kernel subsystem reflects Jupyter's design even though that's not the primary frontend anymore. Factor in .NET Interactive architecture, where each kernel handles rendering, completions, formatting, etc. Starting from the ground up was necessary to avoid the coupling with Jupyter and VS Code.

With a fresh start I was able to decompose the engine by concern rather than by language. In practice, this means a university can build a custom cell renderer for teaching without touching a kernel. A component vendor can ship a data formatter that renders their controls natively in notebook output. A language community only needs to implement one interface to get a working kernel, and everything else (rendering, formatting, magic commands, theming) composes independently. None of these scenarios require forking or modifying the engine itself.