w3forth - make IT small again
FORTH and the art of simplicity
One thing seems to me is gone the last couples of years - simplicity.
There are some programming languages out there which enforces purity in some
concepts. For example SMALLTALK for object orientation, HASKELL for functional
I would say FORTH enforces simplicity. To do simplicity FORTH is not a
requirement, like Smalltalk is not a requirement for writing object oriented
applications. FORTH is an eye opener in this respect.
Before I'm going deeper into detail I would enumerate some fundamental concepts
* FORTH is very small, about some k of size. This mean that there is
not much flesh which have to be maintained and understand.
* FORTH doesn't rely on any operating system but works nicely on top
of most operating systems.
* FORTH is a realtime operating system.
* FORTH contains an interpreter, compiler, editor, debugger, persistence
and multitasker. For an experienced FORTH user this is builds a
remarkable productive system.
* FORTH is self contained, that means that the interpreter, compiler,
debugger, etc. are always available, even at runtime and still requires
only some k of memory. It is convenient to save the current system as
a new executable and start from there or distribute that executable. No
other programs are required (no external compiler or linker).
Using and programming FORTH is brutally simple, but it seems to take years to
master it. Since FORTH is a stack oriented language most developers seems to
have troubles in switching the kind of thinking. A big source of problems arises
if developers try to apply their known features of their current tools to FORTH.
FORTH is working in a very different way to be able to achieve simplicity.
I don't show FORTH code for now, because it is like a foreign spoken language.
It will merely distract the user. So I continue to elaborate the principle
behind the simplicity first.
FORTH is about problem solving. Real problems, not self made problems. It seems
to be difficult to see self made problems, so it is important to focus on the
customers requirement and to ask about the source of the problem which is going
to be solved. For example configuration files, the need for garbage collector,
SQL database, Frameworks and so on.
In writing libraries and frameworks without knowing the problem will end in a
large and less flexible systems which doesn't meet (perfectly) the requirements.
SQL databases are a good example. Because we have to work with SQL we squeeze
the problem into tables, fields and relations. This approach works well, but is
far away from optimal, in my experience about three order of magnitudes.
Therefore I gave up on SQL and switched to a direct memory mapped storage
In reducing complexity we get safety. Another example are files. Files are not
a big thing, but I/O can raise many exceptions (disk full, read or write errors,
file not found, no open-read-write permission etc.). In avoiding files we have
the opportunity to remove 100% of that complexity. Files are a requirement,
how could we abandoned files then ?
One approach is to factor out the reading and writing of files in a different,
independent tiny application. This application transform the files into a
reliable alternative (i.e. fixed size memory map file or any other idea). Or
vice versa, from, for example, memory into files.
But these converter are not part of the business application.
This raises the idea of independent tiny applications. Each application solves
an independent part of the whole application. With this approach the need of
namespaces disappear. The need of dynamic memory management disappears.
At the end of the tiny application all memory will be reseted, a 2 nano second
operation. It should be easy to prove and test tiny applications. Changes
in one tiny application don't interfere with other tiny applications.
But how can we get there ? First we need a deeper understanding of
the requirements and the solution. Writing small prototypes might help here.
Split the requirement into independent parts. UseCase analysis might is
a good approach. Then only implement functionality which is really needed.
05.02.2018, Andreas Klimas - firstname.lastname@example.org