Re: [stack] a concatenative language for the Semantic Web

On 5/21/07, Joseph Huang <[email protected]> wrote:

Hi Joseph!


> What about making / postfix? It seems misleading to have blah op == /blah,
>  especially in a concatenative language.
>
>  How about shortening op to 1 character? Maybe | because it's so narrow. Or
>  just use / instead of op.

I've deliberated a bit about op as a prefix.  It's potentially a
little confusing when you have to match up the text representation
with the RDF representation, but a postfix symbol would not be as
handy for path expressions, e.g.

    foo/bar/quux

...has a pretty obvious path-like quality, whereas

    foo bar op quux op

IMO does not, even if op is replaced by a special character:

    foo bar! quux!

Maybe I'm too attached to infix operators for path expressions, but
the first version looks better to me, and is probably closer to what
someone familiar with Notation3 would expect.


>  How good is Ripple for general purpose computing?

The Java implementation of Ripple is best suited to applications
involving multivalued relations, such as RDF's.  It uses some rather
expensive compositional plumbing to distribute each operation over
arbitrarily many values, so if you don't need that functionality,
you're better off using Joy, Factor, etc.  However, as a
representation language, Ripple can just as well express single-input,
single-output programs, so executing those programs efficiently would
be a matter of writing a specialized evaluator or compiler (also on
the TODO list).  The advantage of using Ripple in that case is that
you could more easily share and re-use programs; rather than
downloading a program manually and then running it, all you need to
execute a Ripple program is its URI.


>  -JSH

All this feedback is great.  Many thanks!

Josh