The road ahead for MirageOS in 2021

For those, such as myself, who were not previously familiar with it:

"MirageOS is a library operating system that constructs unikernels for secure, high-performance network applications across a variety of cloud computing and mobile platforms. Code can be developed on a normal OS such as Linux or MacOS X, and then compiled into a fully-standalone, specialised unikernel that runs under a Xen or KVM hypervisor."

https://mirage.io

It seems an interesting idea.

And here I thought they were referring to MirageOS on the TI-83

https://www.calculatorti.com/os/mirageos/

Great news, specially the deployment in production with products like NetHSM, always great to point out when someone asks where it is used if at all.

https://www.nitrokey.com/products/nethsm

This is really interesting! Congrats on securing grant funding - I can't wait to see what MirageOS becomes.

(Speaking of grants, does anyone know any funding opportunities similar to NGI Pointer but in the US?)

I'll be a bit of a devil advocate here but both the choice of words and attitude are obscuring facts.

First with a name like MirageOS and calling itself a unikernel, you'd think this was an OS. But this actually has no driver, no user-space, no file system. It receives boot parameters and communicates over TCP or UDP. Mmmm I'd call that a program that receives command-line parameters and communicates over the network. But "program" sounds a lot less exciting than "unikernel".

It does not even manage memory or processes or anything like that. It uses a thin C wrapper to load on virtual machines or . Just like you could with any program that does use any OS facilities.

It does not have process, praising cooperative multi-tasking over multi-threading or multi-processing. Cooperative multi-threading means every part of your system is dependent on every other part behaving correctly to yield the CPU at short interval. Mis-design any single algorithm in the system to be quadratic or worse and you can have long, unpredictable scheduling delays. Forget doing anything requiring real-time responses. Even semi-regular response times are not guaranteed under cooperation.

You could achieve the same goals with a statically linked program in any languages...

I really like the idea of unikernels, but I find it a little strange to hamper the performance of one by using Ocaml when performances one of the most significant advantages.