I’ve recently been in a deep discussion exploring various computational models for efficient cloud computing. We began by introspecting Unikernels, which offer single-purpose OS instances tailored for specific applications, resulting in rapid boot times and minimal resource consumption. The discussion moved to considering a hypothetical “bare metal” JVM or BEAM Machine that provides an OS-like abstraction layer for applications. The idea was to have an environment where developers can focus on application logic, while the underlying system efficiently manages hardware resources.<p>This line of thinking brought forth an intriguing proposition: Could Forth, known for its interactive and intimate relationship with hardware, be an innovative foundation for cloud and serverless environments? Here are some focal points:<p>- Interactive System Access: Forth allows for direct memory inspection and modification, potentially streamlining debugging and profiling.<p>- Minimal Attack Surface: With its compact footprint (often <100KB), Forth’s attack surface is reduced, suggesting enhanced security.<p>- Parallel Execution Potential: By statically analyzing the Forth dictionary, one could assign specific words or tasks to individual CPU cores, reminiscent of the thread-per-core approach, and potentially bypass traditional thread management overheads.<p>Given these traits, might Forth be a suitable base for crafting highly efficient, secure, and parallelized systems for specialized cloud applications? Are there pitfalls or challenges I might be overlooking in this approach?<p>Mindful, this will be a new kind of Forth, one built for Big Iron servers, and where HW resource are prime real state, this could be a very interesting approach to hardness more of the potential on “bare metal”.<p>I’m eager to gather the community’s insights on this intriguing prospect.
Back in the days of 8080 CPUs, I built a number of projects in Forth,including a "discovery system" which adapted patterns from the late Douglas Lenat's Eurisko project, as well as a trading system which, on paper, could have been successful.<p>From that experience, I tend to agree with your intuitions.<p>Now that we have 64-bit CPUs, it's tempting to revisit Forth.
What might be missing is he massive collection of libraries we have in other languages.<p>Forth, BTW, historically, has been pushed directly to Silicon in at least two instances, the first one by Chuck Moore [1], the second one designed by Philip Koopman [2]<p>"forth chip" now gives pretty good google.<p>[1] <a href="https://news.ycombinator.com/item?id=8860786">https://news.ycombinator.com/item?id=8860786</a><p>[2] <a href="https://vfxforth.com/flag/jfar/vol6/no1/article1.pdf" rel="nofollow noreferrer">https://vfxforth.com/flag/jfar/vol6/no1/article1.pdf</a>