For Windows systems, I probably would go with either SwiftForth or iForth, again, because the book is (re)written to those implementations.
SwiftForth is $400, and iForth is €100.
SwiftForth is available as a free download. It will lack the source code and other useful libraries, but should be plenty sufficient to learn from:
to have a free evaluation download. However, I no longer see the link.
If you want neither Swift nor iForth, then I'd try to find a Win32 distribution of gforth.
However, again, remember that ANSI's specifications only covers so much. Ultimately, something
in Starting Forth will differ from the ANSI spec, and that's why I encourage using an environment as close as possible to what's listed in the book.
Like I'd said earlier, ANSI Forth is somewhat similar to ANSI C in this regard. People looking to learn C today rarely do so with the expectation that they'll use a VT-100 terminal to access a mainframe; yet, that's about all the standard C libraries cover. The more advanced exploitations of C are rarely considered "standard" (e.g., writing COM objects in C, command-line parsing for DOS or Unix shell commands, etc.), and often requires specific compilers (e.g., each compiler has a unique inline assembly syntax, and its own flavor of the standard C libraries often with embellished interfaces). This is such a problem, in fact, that most open-source tools must explicitly be "pre-configured" via a packaged configure
script. It's not uncommon for the configure script to exceed the compiled program in length!
be transferred, of course; you just have to be observant, but that requires experience, which I'm not assuming the reader has yet. Everything I've said above applies to different Forth implementations too. This is why I advocate sticking with SwiftForth if
the book he's following is also written with SwiftForth users in mind.