Re: So let me get this right: (Was: NFNF vs 1NF ...)

"Mikito Harakiri" <mikharakiri_at_iahu.com> wrote in message news:9RTOd.51$Lh.46_at_news.oracle.com...
> "DBMS_Plumber" <paul_geoffrey_brown_at_yahoo.com> wrote in message
> news:1108083168.253466.60590_at_z14g2000cwz.googlegroups.com...
>> If you can build an XML system that's less than within an order of
>> magnitude of a SQL system, that supports the null hypothesis (it's all
>> in the implementation: not the data model).
>
> Speaking of performance, the usual scenario is:
> 1. implement a thing relationally
> 2. implement it in XML
> 3. discover that relational is 100x times faster
> 4. submit a bug and wait until XML folks investigate
> Maybe it will catch up someday, but with all the language compexity and
> ugliness, do you really believe it?

But here's a little tidbit that I think is quite significant because it isn't just theorizing about possibilities or lack thereof -- the IBM U2 databases, using the old pick model, use almost the same data model as XML and they are FAST! I say "almost" because there are things like attributes vs entities in XML (I hate that!); you can only embed lists a few levels deep in PICK; dtds & xsd are not similar to pick dictionaries that describe rather than prescribe; and I'm sure a few other differences. But it's a short step between those two data models compared to the jumps required for OR mapping. XML & PICK could be seen as two flavors of the same data model.

And did I mention -- PICK IS FAST! At least that's what they tell me and when I was paying somewhere between a penny and a dime to my counterparts' dollars a decade or so ago (the last time I managed a production U2 shop myself), I figured I was saving both in hardware costs and personnel maintaining the software. When I stepped into that situation I found it to be VERY impressive in a big bang for the buck way. It's not quite as impressive today (IMO) with everyone attempting to figure out how to roll their own GUI.

So, RDBMS users can laugh at the current performance of XML databases (I know I have ;-) but the data model itself is not what's slowin' em down, at least not those aspects of the data model that are the same as PICK. I'm thinking someone will figure it out soon enough. I don't think it will be long before people will realize that if they can stick with the XML data model throughout their applications, they can save dollars. We will then be talking about having to interface with those legacy RDBMS systems. smiles. --dawn Received on Fri Feb 11 2005 - 04:19:50 CET