> In the FAQ.[txt|html], there is the following code. I'm pretty sure
> I have no idea why the advice is given to use a BufferedInputStream.
> First of all because we are reading into a byte array.
>
> Second, I'm not sure why one should "ever" use a BufferedInputStream
> since it appears to me that the result of #getBinaryStream is
> already buffered.
>
> What am I missing?

I think there's no point in using BufferedInputStream. You're right,
driver buffers parts of data (size of the buffer is specified in
blob_buffer_size parameter, default, afair, 16k) and using
BufferedInpurStream only adds an unneeded overhead. Probably the only
reason to use BufferedInputStream are mark()/reset() methods when
using big in-memory buffer, so that seeking happens locally, not remotely.

Roman