Re: [Info-Ingres] NoSQL - the end of rdbms?

On Mar 2, 2010, at 8:39 AM, Paul White wrote:

> Couple of interesting articles on NoSQL this week
>
> NoSQL in the real world
> http://news.cnet.com/8301-13846_3-10451248-62.html

Gah. Where is Fabian Pascal when you need him?

It's painful to watch a hierarchical data model being touted as if
it were something new, just as it's painful to see the conflation of
"relational" and "SQL" as if they were the same thing. I got a couple
paragraphs in and gave up.

I guess that if you have a situation that is best modeled
as a hierarchy, a hierarchical DBMS will handle it the best.
IMS, anyone? Of course, the hierarchical model
struggles mightily with anything that isn't naturally a tree.

Hammers are great for driving nails, not so good at
turning screws...

Karl

Karl Schendel wrote:

>
> On Mar 2, 2010, at 8:39 AM, Paul White wrote:
>
>> Couple of interesting articles on NoSQL this week
>>
>> NoSQL in the real world
>> http://news.cnet.com/8301-13846_3-10451248-62.html
>
> Gah. Where is Fabian Pascal when you need him?

Unfortunately he exploded in a moist cloud of logic, frustration, and
idealism. Poor man. (He was last seen writing a political blog; he is
an economist by training.)

> It's painful to watch a hierarchical data model being touted as if
> it were something new, just as it's painful to see the conflation of
> "relational" and "SQL" as if they were the same thing. I got a couple
> paragraphs in and gave up.

Yep, drives me nuts too. Just because SQL sucks so much your ears pop
doesn't make set theory and predicate logic the wrong tools for the
job. Set theory has been around for a few hundred years and predicate
logic for a couple of thousand. They haven't needed an upgrade for
quite a while. "No" to SQL for sure; but there is no credible
alternative to the relational model. We just need someone to implement
it properly and devise a decent query language (or two or three--I don't
care how many).

> I guess that if you have a situation that is best modeled
> as a hierarchy, a hierarchical DBMS will handle it the best.

I can only assume you've never tried representing trees of arbitrary
depth in hierarchical database (or an XML document). It can't be done
any easier than with a table. In fact in the case of XML, it can't be
done at all except by representing it as a table. The parse tree isn't
the data.

--
Roy

UK Ingres User Association Conference 2010 will be on Tuesday June 8 2010
Go to http://www.iua.org.uk/join to get on the mailing list.

On Mar 2, 2010, at 9:27 AM, Roy Hann wrote:

> Karl Schendel wrote:
>> I guess that if you have a situation that is best modeled
>> as a hierarchy, a hierarchical DBMS will handle it the best.
>
> I can only assume you've never tried representing trees of arbitrary
> depth in hierarchical database (or an XML document). It can't be done
> any easier than with a table. In fact in the case of XML, it can't be
> done at all except by representing it as a table. The parse tree isn't
> the data.

I will admit to never having used any hierarchical DBMS
at all. I never got close enough to mainframes to
experience IMS, and XML ... well, never mind.

It's just annoying to read pieces that say "problem X ran
faster on implementation A than it did on implementation B.
Therefore A's data model is great and B's is obsolete."
I mean, can we even get some basic logic going here, please?

Karl

[rajus01]

It is good that you brought this topic up Paul...

Recently I was reading about Twitter is moving to Cassandra from MySQL.... I was also wondering about the same thing...

It is interesting to see Karl's and Roy's thoughts on this topic of NoSQL databases....

Cassandra @ Twitter: An Interview with Ryan King � MyNoSQL

Regards,
Usha

Ingres Forums wrote:

> It is good that you brought this topic up Paul...
>
> Recently I was reading about Twitter is moving to Cassandra from
> MySQL.... I was also wondering about the same thing...
>
> It is interesting to see Karl's and Roy's thoughts on this topic of
> NoSQL databases....

Then allow me to elaborate a little more. :-)

Much of the NoSQL movement is about solving problems that don't
benefit from--or actually suffer from--the features an SQL DBMS
provides.

That is totally cool. SQL DBMSs are actually intended to *prevent* a lot
of things. If you read some of the more technical books you will see
that they are mostly concerned with constraints and consistency--with
stopping things that are considered adverse for a well-defined purpose.

If you have a different purpose and you want to do the very things the
SQL DBMS is *actively* trying to prevent, obviously you shouldn't use
one.

I, on the other hand, am concerned exclusively with what SQL does
try to do, and I wish it did it more and did it better. NoSQL-type
solutions offer me no help at all.

--
Roy

UK Ingres User Association Conference 2010 will be on Tuesday June 8 2010
Go to http://www.iua.org.uk/join to get on the mailing list.