Ask HN: How much would it cost to create a new development stack?

[lukev]

I'm entirely sold on the idea of VMs. The idea of multiple languages compiling to a target bytecode running on a cross-platform, high-performance VM is in my opinion one of the biggest advances in software development in the last two decades.

The CLR is technically superior, but is not really free, and not entirely cross platform.

But with Oracle starting to clamp down on Java, it occurs to me that there's no truly unencumbered VM out there.

Is now a good opportunity for someone to step in and build a third option, completely open, incorporating lessons learned from both the CLR and the JVM? Of course, it would be an absolutely monumental task - but not impossible. How many developers would it take? Maybe 50, working for 2-3 years, to get a usable first cut and basic standard libraries? Paying them an average salary of 100k, that's "only" 15 million dollars. Not to mention all the open source contributions - this is the type of project that could attract OS contributors like moths to a flame.

$15 million is a lot, but its negligible compared to the economic and cultural impact of a totally open VM superior to both the JVM and the CLR. It seems well within reach, if funded by a non-profit trust established by a large number of invested firms. Hell, there's probably a few individuals wealthy enough to pull it off, and who are philosophically inclined to support such a project.

Is there something wrong with my reasoning, here? Do you think there's any chance we might actually see a project like this take off? Would you follow/contribute to such a project?

[wmf]

I would expect the cost of building a self-sustaining ecosystem (including libraries, books, conferences, etc.) to be 10x the cost of building the platform itself. Having a charismatic leader like DHH or Shuttleworth would allow much of the ecosystem-building to be offloaded onto the community.

[stonemetal]

LLVM, Squeak, Erlang, CPython, pypy, Rubinius, parrot, several of the javascript VMs all "truly unencumbered VM". If I were going to start a massive library so everyone wants to use my VM project today I would start with either LLVM, parrot or pypy probably in that order.

[spooneybarger → stonemetal]

the squeak vm will probably be replaced at some point by cog: http://www.mirandabanda.org/cogblog/

[drlisp]

>>Maybe 50, working for 2-3 years, to get a usable first cut and basic standard libraries? Paying them an average salary of 100k, that's "only" 15 million dollars.

I think you are underestimating the man-hours and cost. You will need a Google or some other tech giant to engineer this. Why not improve openJDK?

[spooneybarger → drlisp]

correct me if i'm wrong but from what i've read about the patents on this, some improvements might open you up to patent issues.

[spooneybarger]

While slow right now... parrot shows some promise: http://en.wikipedia.org/wiki/Parrot_virtual_machine

[lukev → spooneybarger]

Interesting. Do you know what the state of Parrot libraries is? (It's not evident on their website). Unfortunately, a VM/environment has to be bootstrapped in this regard - nobody wants to write their own socket library just to build a webapp.

[spooneybarger → lukev]

i haven't been following close enough to know. if someone really wanted a new stack, the parrot vm is just what immediately jumps into my mind as it is designed for running multiple languages and having them interact.