Commentary on Aaron Weiker's article and Richard Hale Shaw (C# MVP) blogs on "Moving away from C++"

C# is not acceptable in some situations for high performance mathematical code. E.g. In the forthcoming SQL Server 2005 Data Mining release, you can write native data mining algorithms in C++ but not so natively in C#. I was disturbed about this and I asked the good folks at Redmond what was going on, since, the presentation seems to be that C# is the way to go.

The response was that that C# is not a first class candidate for algorithmic development targeted at perhaps billions/trillions of records and several columns or attributes of data such as you will find in data mining.

I was a bit baffled because at the same time, MS is positioning C#, VB.NET and .NET Framework as a first class development language on SQL Server 2005 while advising the community to move away from using C++ in SQL Server 2000 development such as in extended stored procedures.

My position is that C++ will always be needed where a lot of data is being analyzed such as in SQL Server 20005 Data Mining, but, even though C++ will outperform C# in such instances, clients and developers more often will use C# because of the productivity tradeoff.