Re: [Axiom-developer] Project directions

[Stephen Wilson]

Hi Ralf,

I think you have made an excellent observation.

Ralf Hemmecke <address@hidden> writes:
[...]
> with a clear specification it doesn't matter whether the compiler is
> written in Boot, SPAD, LISP, or C or Haskell. The compiler should be
> called if needed. Why would the language it is written in matter?

One of the main problems is simply getting all of the different
components to play well together.  This basically means defining a
representation for a byte stream which can be passed back and fourth
from one application to the next.  No real difference conceptually
than an internet protocol on the one hand and a file format on the
other.  The big issue is that the byte stream (what ever that may
be. There would in every likely hood be many sorts) be interpreted
efficiently and universally by all system components.

Using the same programming language to implement them all alleviates
that problem significantly since everything is internal data.  Thus,
the former approach introduces real engineering problems.  

[...]
> Why cannot that all live under the hat of "Axiom"? Some people like to
> have a compiler in LISP, some in BOOT. Those people don't agree, which
> will probably lead to two compilers that do the same thing. It is only
> inportant that the specification is clear. Would it be bad to have two
> compilers? Three?

Fundamentally however, there is no problem with this in principle.  Of
course having everyone cooperate on realizing compatible
interpreters/compilers is a much bigger challenge than simply agreeing
on what the proper programming language to use is.

It is also difficult (and not nearly as enjoyable, which is important
in open source work) to write systems which are trying to reach new
horizons when you constrain fundamental interactions to byte-streams
:)

> Clearly, if people don't agree on one language that splits developers,
> but the goal is to provides a mathematical and intuitive user
> interface that is supported by a typed programming language. We all
> work to that goal don't we? There is no need to fork again just
> because some people like LISP and some don't. The common goal lives
> some levels up.
> 
> Collaboration, not separation.

I could not agree more.  It does not matter to the user -- whose
opinion in the end is what matters most -- what language the system is
implemented in.  Unfortunately some of us prefer different approaches
to meeting the expectations.  I personally see no problem with having
different groups of developers pursue common goals along different
paths, even if it does divide an already scarce resource.  Too many of
the ideas, designs, concepts, etc, transcend the programming language
used.  Rich collaboration in any event is certainly possible.


Sincerely,
Steve