Re: Hints on creating a .NET driver

Sven Köhler wrote:

>> 2) There exists now a C++ API that will form the basis of future 
>> implementations of the Precompiler, ODBC and possibly a less portable, 
>> but faster JDBC driver. The sources are in 
>> sys/src/SAPDB/Interfaces/Runtime and sys/src/SAPDB/Interfaces/SQLDBC, 
>> documented by doxygen comments (sys/config/interfaces/doxygen).
> 
> 
> Well, every bit of native code makes a program more instable. I was 
> abled to crash the Sun's zlib implementation for example. It is used to 
> access any JAR and ZIP file with Java. The whole VM crashed and the main 
> reason why i use Java (a Java program cannot crash) seemed to be absurd.
 >
 > I highly recomm to have 100% Java/.NET implementation and a optional
 > speed optimized native part if possible.

I don't know how much of my Java experience translates to .NET:
- The first version of the JDBC driver was significant slower than the 
other interfaces because filling structures byte by byte is simply slow 
in Java
- calls to native methods were very slow, so implementing only parts as 
native methods didn't seem worthwhile. I suspect that the .NET 
equivalent of JNI is much faster, but will also require strange rituals 
involving human sacrifice to prevent the whole program from blowing up.


Daniel Dittmar

-- 
Daniel Dittmar
SAP DB, SAP Labs Berlin
[email protected]
http://www.sapdb.org