 | | From: | Mok-Kong Shen | | Subject: | New relational techniques? | | Date: | Wed, 12 Jan 2005 19:44:20 +0100 |
|
|
 |
I saw a site http://www.pilesys.com/tech.htm claiming to have
developed revolutionary techniques of representing relations
and computing with them. It would be fine, if some experts
would say their opinions about them and eventually give a
concise and more easily comprehensible summary of the materials.
(Of the papers there the one entitled 'Order in complex networks'
seems to be of more relevance than the rest.)
Thanks.
M. K. Shen
|
|
| | From: | Ed Prochak | | Subject: | Re: New relational techniques? | | Date: | 14 Jan 2005 05:18:56 -0800 |
|
|
| Here's a little deeper impression:
They want to reduce "redundancy" of things like the alphabet. (ie
storing the letter A only once). But they never consider the
implimentation costs (how many bytes in the links between their
socalled "objects". Do you really save anything if it takes two 32bit
pointers to reference a single letter?) The description reminds me of
an exercise from a Data Structures class to be able to load a list and
report it back, sorted in two different ways.
They have a paper called "Pile - A Mathematician's View" which is
merely a longer this-is-great quote from an employee, Polina
Proutskova. There is no mathematics in that paper at all. Where's the
mathematical foundation Polina?
I love this from one of their White papers. Really techynical
presentaions:
"But while the game dynamically computes these frames on demand, and
thus can generate zillions of different versions "
How many decimal places in 1zillion? Can anyone tell me?
Finally, you can tell it is all just marketting so far, when you browse
to the PRODUCTS page. There you find....Nothing. Well, only an offer
for a Test and Evaluation package and a information about a user group.
tell me, this is an early April Fools project isn't it? No really this
has got to be a spoof of Neo's systems and others like it. All the web
site has is promises of great technology but no implementation. If
there is something better, show us.
Just my current view. I can be convinced, but not by lists of quotes.
The only one that seemed appropriate was this one:
"...certainly a major breakthrough if claims stand up..."
Joel Birnbaum, former director HP Labs, Palo Alto
So I will wait for a domonstration of a working system. 'til then I
remain very skeptical.
Ed.
|
|
| | From: | Gene Wirchenko | | Subject: | Re: New relational techniques? | | Date: | Fri, 14 Jan 2005 09:36:51 -0800 |
|
|
| On 14 Jan 2005 05:18:56 -0800, "Ed Prochak"
wrote:
[snip]
>I love this from one of their White papers. Really techynical
>presentaions:
>"But while the game dynamically computes these frames on demand, and
>thus can generate zillions of different versions "
>How many decimal places in 1zillion? Can anyone tell me?
I can. Since zillions is obviously a count of the number of
versions, it is an integer and thus has zero decimal places.
>Finally, you can tell it is all just marketting so far, when you browse
>to the PRODUCTS page. There you find....Nothing. Well, only an offer
>for a Test and Evaluation package and a information about a user group.
>
>tell me, this is an early April Fools project isn't it? No really this
>has got to be a spoof of Neo's systems and others like it. All the web
^^^^^^^^^^^^^^
I do not like it. A nice bit of ambiguity there.
>site has is promises of great technology but no implementation. If
>there is something better, show us.
>
>Just my current view. I can be convinced, but not by lists of quotes.
>The only one that seemed appropriate was this one:
>"...certainly a major breakthrough if claims stand up..."
>Joel Birnbaum, former director HP Labs, Palo Alto
Tee-hee.
>So I will wait for a domonstration of a working system. 'til then I
>remain very skeptical.
But it is the latest and greatest!
Sincerely,
Gene Wirchenko
|
|
| | From: | Neo | | Subject: | Re: New relational techniques? | | Date: | 14 Jan 2005 10:38:10 -0800 |
|
|
| > ... this is an early April Fools project isn't it? No really this
> has got to be a spoof of Neo's systems and others like it...
> So I will wait for a domonstration of a working system.
> 'til then I remain very skeptical.
// Create items in directory to classify things
(CREATE *person.item ~in = dir)
(CREATE *student.item ~in = dir)
(CREATE *dba.item ~in = dir)
// Create student Mok and dba Ed
(CREATE *mok.cls = person)
(CREATE mok.cls = student)
// Create DBA Ed
(CREATE *ed.cls = person)
(CREATE ed.cls = dba)
// Create data models
(CREATE *data model.item ~in = dir)
(CREATE *pile.cls = data model)
(CREATE *rm.cls = data model)
// Create verbs
(CREATE *study.cls = verb)
(CREATE *disbelieve.cls = verb)
(CREATE *trust.cls = verb)
(CREATE *advise.cls = verb)
(CREATE *because.cls = verb)
// Create relations
(CREATE mok.study = pile)
(CREATE ed.disbelieve = pile)
(CREATE ed.trust = rm)
(CREATE ed.advise = mok)
(CREATE (ed.disbelieve=pile).because = (ed.trust=rm))
(CREATE (ed.advise=mok).because = (mok.study=pile))
// Find a dba who advises a student who studies a data model
// Finds Ed.
(SELECT %.cls=dba
& %.advise=(%.cls=student
& %.study=(%.cls=data model)))
|
|
| | From: | Ed Prochak | | Subject: | Re: New relational techniques? | | Date: | 12 Jan 2005 14:51:07 -0800 |
|
|
| given that the site opens complaining about "our current paradigm of
redundantly representing and storing data in predetermined hierarchical
container structures" sounds like a rage against network model
databases of 20years ago.
First impression is that it is yet another DBMS wannabe.
but I'm no "expert", just a very seasoned developer.
Ed
|
|
| | From: | David Schofield | | Subject: | Re: New relational techniques? | | Date: | 15 Jan 2005 03:30:06 -0600 |
|
|
| On 12 Jan 2005 14:51:07 -0800, "Ed Prochak"
wrote:
>given that the site opens complaining about "our current paradigm of
>redundantly representing and storing data in predetermined hierarchical
>container structures" sounds like a rage against network model
>databases of 20years ago.
Or against XML???
David
>First impression is that it is yet another DBMS wannabe.
>but I'm no "expert", just a very seasoned developer.
>
> Ed
>
|
|
| | From: | Neo | | Subject: | Re: New relational techniques? | | Date: | 12 Jan 2005 19:52:19 -0800 |
|
|
| > It would be fine, if some experts would say their opinions
> about them and eventually give a concise
> and more easily comprehensible summary of the materials.
I am no expert, but below script for experimental db shows how to
create relations that can be referred in other relations and so on as
shown in Fig 2 & 3. It can also be accomplished in RMDBs by using
generic modelling.
// Create persons John and Mary.
(CREATE *person.item ~in = dir)
(CREATE *john.cls = person)
(CREATE *mary.cls = person)
// Create verbs: like and because.
(CREATE *like.cls = verb)
(CREATE *because.cls = verb)
// Create initial relations.
(CREATE john.like = mary)
(CREATE mary.like = john)
// Here the above relations are referred in a new relation
// and this process can be repeated.
(CREATE (john.like=mary).because=(mary.like=john))
// An example which repeats the above process 4 levels deep
// is shown at www.xdb2.com/Example/Ex136.asp
|
|
