Re: [xml-dev] functional databases noodling ...

[James Fuller <james.fuller.2007@gmail.com>]

On Mon, Feb 20, 2012 at 9:57 PM, Jim Tivy <jimt@bluestream.com> wrote:
> Thanks all for bringing my attention to IRIS.
> I think the idea of database starts with data model.
> Then there needs to be a query language for that data model and that the
> query language is closed with respect to the data model.
> The idea of "functional database" can mean a number of things.
> 1. that functions are first class parts of the data model.
> 2. That the query language is functional.
>
> Perhaps another way to look at "functional database" would be to declare
> that side effects are not possible.  Read Repeatability certainly has that
> and snapshot semantics address that as well.

I agree that the data model is where to start but I wonder if the impl
doesn't have a say ... e.g. would it follow that  MVCC could be
considered (in isolation and ignorance of higher abstractions on the
database) as a 'functional' characteristic of a database ?

as for side effects, defining what everyone means e.g. level of purity
(of side effects) can be problematic ... most functional programming
languages make some minor compromises along these lines.

from Dr. Kay's responses, the impression I got was the idea that every
'fact' in a functional database would have one representation or by
dint of referential transparency all representations would boil down
to the same immutable value ... there are some interesting subtleties
involved with that.

to me what sits on top of the database in terms of query language
should be discussed after we agree (or more like educate me!) on what
comprises the lower levels of a functional database

Jim Fuller