Re: [PhillyOnRails] jruby + hadoop?

[Toby DiPasquale <toby@cbcg.net>]

On Thu, Sep 06, 2007 at 03:32:27PM -0400, Mort Goldman wrote:
> Interesting conversation and I look forward to the talk.  One historical
> question.  Wasn't the intent of Codd going back to "Large Shared Data Banks"
> to avoid the pain of "stateful" implementations of schemas and access
> methods?  Is the resultant need for compiled SQL queries and complex RDBMS
> administration to achieve performance goals a failure of relational theory,
> a failure of RDBMS implementations (including the market decision to adopt
> SQL vs. using a relational calculus based language like QUEL), or a
> migration to a different problem domain (semi-structured vs. structured
> data).

I believe, personally, that its largely the latter. For the kinds of
things people tend to want to do these days, semi-structured data is the
best you can get (e.g. the Web, email, etc) Therefore, the absolute
structure of the RDBMS model is at an impedance mismatch with the data you
want to work with itself. Basically, we no longer live in a world where
the data you need to work with fits nicely into a set of pre-determined
columnar attributes.

You can apply relational set algebra to semi-structured data just fine
(see the Pig research language from Yahoo!) so I wouldn't necessarily say
that its a failing of SQL or relational theory. I can't really speak to
Ted's original intent, though, but its a very good question to ask.

-- 
Toby DiPasquale
_______________________________________________
To unsubscribe or change your settings, visit:
http://lists.phillyonrails.org/mailman/listinfo/talk