ON A PILE OF ... WHAT?

From: Ralf Barkow

Date: 02 Feb 2006

I feel, you might want to take a look at the truly relational Pile technology at and our open source project PileWorks. Looking forward to discuss our approach with you.

From: Fabian Pascal

How can something truly relational be expressed as follows?

Computing today is mainly data processing in data spaces, where objects and events of the 'real world' are represented by physical data organized in hierarchical structures.

From: Ralf Barkow

Thanks for the reply! I'm sorry that the very first sentence on our website might get you to turn away. You are absolutely right, that we Pile Guys have some problems with the language!? That is the really strange language of the inventor of Pile, Erez Elul.)

You might give us a second chance, reading the second sentence: Pile opens the door to a new [for us, maybe known for you!] computing paradigm: Here objects and events of the 'real world' are assimilated into a non-representational relation space, where data are dynamically generated ... and please have a look at the attached paper: The Pile System: A New Approach To Data And Computing (Polina Proutskova)

Written by a mathematician, Pile is introduced from a graph perspective.

From: Fabian Pascal

Read Hugh's reply. To be honest, I don't think that all who are involved with Pile know the relational model at all, or if they do, they don't understand it. The chance for them to have developed a truly relational system is zero.

From: Hugh Darwen

It's not clear to me what the claim is. I'm only interested in claims of conformance to the relational model of data, preferably as expressed in The Third Manifesto. If no such claim s being contemplated, then what I observe briefly here is irrelevant; otherwise, the authors could do us all a favour by using the number points of the Manifesto to justify their claim in a way that would allow others to validate it.

The documentation seems impenetrable (and some of it physically unreadable in Mozilla Firefox because for some reason it doesn't word wrap). But I discovered this:

... the Pile System implements a completely new approach of data representation.

If that's true, I cannot see how it can claim to be truly relational in our sense of that term, because truly relational is 36 years old.

I also spotted:

All these features are gained due to generically converting the implicit use of the traditional pointer to a "knowing reference.

That looks like The Second Great Blunder to me, and therefore a show-stopper.

Nodes and handles look suspiciously like extraneous concepts too—unless they can be equated to the normal terms used in relational database theory. (I am making no attempt at this time to understand what they really mean—just glancing at the words I see.)

Sorry, but any attempt by me to study the existing documentation further is doomed to failure, so I shan't try. I'm not expressing an opinion on the actual concepts of Pile here, because I don't know what they are.

Posted 3/24/06