+1 to what BJ says.
When the signal has come through your soundcards A/D converter, it doesn't matter anymore if you compress, say, inside the soundcard or in SONAR, because anyway you'll have adjust the recording level in the input of the soundcard. That means, if you put the input level low enough so signal doesn't clip, but reaches peaks of -12 to - 6 dB, that level is good for recording, assuming the singers mic technique is any good. Compressing after that (but before SONAR) does not make the recorded audio any more usable than simply by putting the compressor in the FX bin.
OTOH, I have often used the compressor in my EMU 1820 interface for convenience. I guess anything's OK if you know what's happening, ie. don't have false expectations.