[foaf-protocols] EAV

Kingsley Idehen kidehen at openlinksw.com
Thu Dec 16 23:28:18 CET 2010

On 12/16/10 4:39 PM, Nathan wrote:
> Kingsley Idehen wrote:
>> The key issue here is "Reference" values in the DBMS. Relational DBMS 
>> engines traditionally didn't allow this (until ORDBMS engines came on 
>> the scene but this was done in none standard fashsion).
>> To conclude, EAV + URIs is old. There's a massive audience out there 
>> that doesn't make this connection with RDF, just as the tiny RDF 
>> audience doesn't generally make this connection with the DBMS world.
>> ...
>> EAV (the data model) has never been a subset of RDF, it predates RDF. 
>> De-referencable Identifiers in DBMS engines (Object and Object 
>> Relational) also predates RDF and the Linked Data meme. What RDF 
>> delivers is an optional mechanism for doing for Data what HTML did 
>> for Documents re. hypermedia.
> Yup I follow, I'm just interested in whether it's formally the 
> definition that EAV == RDF, or whether the correlation (which I also 
> see) is one you have personally made (which doesn't undermine it any 
> way).

My correlation based on my travels through many worlds of data access, 
integration, and management. Basically, a lot of what is expressed in 
the architecture of our products. Our ability to make Linked Data from a 
vast array of data source types is code for: Look at what EAV can do for 
you :-)

> As you say yourself, URIs in RDF don't have to be 
> dereferenceable/addresses, they're merely logical constants, names, 
> thus how can they correlate to dereferencable identifies in dbms engines?

Maybe I wasn't clear. I meant: you can de-reference a Reference value in 
a DBMS table. Illustra (work by Prof. Stonebraker based on Postgres pre. 
PostgreSQL) is an example of an ORDBMS that basically did this with 
extended SQL. Even more relevant, myself, the esteemed Prof. and Orri 
(Virtuoso Program Mgr. & Lead Developr) has some intense debates about 
this matter re. RDF and the next frontier for DBMS tech circa. 2000. I 
told the prof. I was going to bet the house on what the W3C was doing 
with RDF and OWL. And of course, he's received periodic reminders from 
me as Virtuoso executed on exactly what we articulated to him way back 
in 2000. On his front, the column-store approach to DBMS data management 
is where he placed his focus, and very soon you'll officially see the 
fusion of this old journey.

A URI can be a Name or an Address. When a Name resolves to an Address it 
doesn't make it an Address.
An Address can be use to unintuitively assign Names to Entities (Things).
There's a delicate nuance in the statements above.

Resolvable URIs via schemes such as HTTP  add network expanse to the old 
concept of dereferencable identifiers. That's the great ingenuity the 
URI abstraction. It's also the part most don't totally grasp. HTTP based 
URIs lets us pull off old tricks in newer and much broader context i.e, 
InterWeb. Ultimately, it will even get people to understand how pointers 
work, since we have a generation of programmers who don't even know what 
pointers are, post emergence of Java.

HTTP URIs have delivered the long desire infrastructure for programming 
language, operating system, application independent distributed data 
objects. Linked Data is classic example of such power. WebID is a great 
example of Linked Data prowess and so on re. this continuum.

> I'm aware that "linked data" implies the dereferenceability in the 5* 
> system, but there's certainly nothing formal, and no semantics in any 
> of the mediatypes to indicate any kind of hyper-ness (other than non 
> normative notes in N3) you won't find any mention of "dereference" or 
> using URIs as anything other than names in any of the RDF specs or 
> serializations.
> Thus, I can see how linked data could do for RDF what HTML did for 
> documents, and how other types of data could be viewed as linked data 
> with certain glasses on, and how all of that could be stored in EAV 
> based DBMS systems - but ultimately I can't see how there's any EAV == 
> RDF, whilst I can see that ( linked data is like webized EAV and EAV 
> compatible )


RDF Schema != RDF.

It's all about 3-tuples as basis for FOL in the data itself. Basically, 
structured data that can be reasoner / inference palatable.

OWL is a strand of FOL.

Javascript != Java.

.... list goes on....

The problem is that RDF is fundamentally about a markup language. Just 
as HTML is a markup language. Pull back encodings and you are going to 
find EAV.

BTW - What's a Windows INI file's underlying model? What about the 
Registry (Hive)? What about those old Windows Help Files of yore? I 
haven't even hit NeXT, Apple File Manager, Hypercard, and raft of other 
issues. At the base you will find a common EAV pattern re. data 
structures. If RDF wasn't EAV how could we make a scalable quad store 
using an RDBMS engine based on a single table with 4-columns? How could 
we make RDF Views using same DBMS?

> Perhaps not worth labouring the point - last thing either of us wants 
> is more confusion when we seek simple clarity.

Put RDF to one side, it cannot be pitched as being everything. Just as 
everything isn't a Resource. We exist in a continuum, lets fit into said 
continuum and lots of inclusion induced progress will be the reward.

The RDF narrative is wrong and broken. If this weren't true, then why on 
earth does it trigger such hard feelings wherever it shows up?


1. http://ycmi.med.yale.edu/nadkarni/eav_cr_frame.htm -- and older EAV note
-- just found this via a quick Google lookup .

> Best,
> Nathan



Kingsley Idehen	
President&  CEO
OpenLink Software
Web: http://www.openlinksw.com
Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca: kidehen

More information about the foaf-protocols mailing list