What's the compelling argument for Wasm (and Boxer) for a Wasm skeptic? It seems to be that a lot of development to create a new environment to develop for that is different than the interfaces we've been developing against for 40 years without an obvious improvement. And we have to re-develop our language implementations to target Wasm. It just seems like a lot of work for a very narrow improvement.
I see it more as moving the browser into being like all the other development platforms, with a complete ecosystem of languages that have different benefits and drawbacks. Forcing everything through javascript has been problematic.
At the bottom WASM is 'just' another ISA, e.g. compilers just need to add a new backend, while interfaces / APIs / languages don't need to change (or just as much as switching between CPU archs like x86 vs ARM).
Main difference (and main advantage) to other ISAs and VM bytecodes is probably that WASM has been built from the ground up with the requirement to run untrusted code safely (because that is absolutely needed for the web browser use case) while not 'leaking' any safety-related requirements up into the programming languages.
Anything on top of the 'WASM is just another ISA' idea is mostly just the usual Silicon Valley hype machine at work.
wasm gives you nothing more than any modern kernel. in fact, after you add the extensions required for your program to do literary anything, like accessing fs, network, etc, then you're out of wasm sandbox and back into depending on the kernel safeguarding you from the wasm host anyway. so what's the point?
This kind of reminds me of the days when Java™ was popular. I agree, if we're pushing to build software to target another machine (WASI) have we really improved anything?
It's much easier to sandbox in theory. We need an easy on ramp to get existing software to run, but once things are more comfortable being written wasm first then we can really see the net improvements.
I always thought this was WASM "value add" -- The "virtual machine" of the browser (although we had this back in the day with Shockwave, Flash and Java applets too hmmm :D)
Yeah, I guess Sun really missed the opportunity to showcase a Java applet as the 'model/controller' for a web page 'view'. Instead we just got applets as a little window in a big window.
What's the compelling argument for Wasm (and Boxer) for a Wasm skeptic? It seems to be that a lot of development to create a new environment to develop for that is different than the interfaces we've been developing against for 40 years without an obvious improvement. And we have to re-develop our language implementations to target Wasm. It just seems like a lot of work for a very narrow improvement.
Not sure how dependent on wasm as such this is, but sub-millisecond cold-starts[1] seems like a pretty compelling argument
[1] https://opensource.microsoft.com/blog/2025/03/26/hyperlight-...
I see it more as moving the browser into being like all the other development platforms, with a complete ecosystem of languages that have different benefits and drawbacks. Forcing everything through javascript has been problematic.
At the bottom WASM is 'just' another ISA, e.g. compilers just need to add a new backend, while interfaces / APIs / languages don't need to change (or just as much as switching between CPU archs like x86 vs ARM).
Main difference (and main advantage) to other ISAs and VM bytecodes is probably that WASM has been built from the ground up with the requirement to run untrusted code safely (because that is absolutely needed for the web browser use case) while not 'leaking' any safety-related requirements up into the programming languages.
Anything on top of the 'WASM is just another ISA' idea is mostly just the usual Silicon Valley hype machine at work.
that's buying the marketing.
wasm gives you nothing more than any modern kernel. in fact, after you add the extensions required for your program to do literary anything, like accessing fs, network, etc, then you're out of wasm sandbox and back into depending on the kernel safeguarding you from the wasm host anyway. so what's the point?
I thought WASM allowed capability based permissions for the code running inside, or has that fallen by the wayside, which was my sad expectation?
It would be more interesting if they made WASM into more of a mini-OS, like BEAM.
That is clearly the logical conclusion.
This kind of reminds me of the days when Java™ was popular. I agree, if we're pushing to build software to target another machine (WASI) have we really improved anything?
It's much easier to sandbox in theory. We need an easy on ramp to get existing software to run, but once things are more comfortable being written wasm first then we can really see the net improvements.
I find WASM compelling as a way to improve browser-based applications without needing to touch any Javascript frameworks.
I always thought this was WASM "value add" -- The "virtual machine" of the browser (although we had this back in the day with Shockwave, Flash and Java applets too hmmm :D)
Yeah, I guess Sun really missed the opportunity to showcase a Java applet as the 'model/controller' for a web page 'view'. Instead we just got applets as a little window in a big window.
GitHub repo: https://github.com/dphilla/boxer
lol. the arguments are all against everyone using "from Ubuntu" and then show how much better they are with a "from scratch" example.
I mean if they don't interoperate with docker as a deployment vehicle the project is pretty much DOA.