Re: Code in the database or middle tier (the CLR controversy)

SAN3141_at_netscape.net wrote:
> There doesn't seem to be consensus about when to put code in the
> database or in the middle tier. There was a long discussion about this
> in an Oracle newsgroup (message ID:
> ULcQb.466$KU5.37_at_nwrddc02.gnilink.net).
>
> Elsewhere there's been discussion about Microsoft SQL Server 2005
> adding the CLR to support stored procedures in languages such as C#. A
> scan of the Web and discussion forums finds differing opinions about
> this.
>
> Two authors have written articles that fall on different sides of the
> debate.
>
> "Keys to the Database"
> http://www.intelligententerprise.com/showArticle.jhtml?articleID=50500830
>
> "SOA, Multi-Tier Architectures and Logic in the Database"
> http://www.sqlsummit.com/Articles/LogicInTheDatabase.HTM
>
> Joe Celko wrote the first article, but his objections point to
> Microsoft SQL Server 2005:
>
> "I have an article at WWSUG.com on how much I hate the CLR stuff that
> Microsoft is putting out."
> http://blog.intelligententerprise.com/archives/002419.html
>
> "The bad news is that SQL Server 2005 will you define your own
> aggregate
> functions in a CLR language."
> Message id: 410d9a51.0502190442.bd68cbe_at_posting.google.com
>
> IBM DB2 and Oracle are doing the same thing with the .NET CLR. Is this
> a non-issue or are all three companies misguided?
>
I don't see what one has to do with the other. First: CLR is a piece of technology. It's support for procedures and functions allows for it's placement on the server. It doesn't force it anymore than PL/SQL, Java, C, Cobol, .... Second: User defined aggregates are an extensibility option for DBMS. _When_ the function is needed then it is better placed in the engine because the DBMS can parallelize the aggregation in both SMP and MPP. TEh App cann not this and not placing teh aggregation in the engine will swamp the network.

Cheers
Serge

-- 
Serge Rielau
DB2 SQL Compiler Development
IBM Toronto Lab