No, there certainly are levels. Lossless just means that if you compress
a picture, then uncompress it, the result is byte-for-byte (or
pixel-for-pixel) identical to the input. It says nothing about how much
smaller the compressed is compared to the original. Of course, with
lossless compression, information theory says there's going to be a
limit to the minimum size of the compressed file. However, at least by
default, System.Drawing.Imaging comes nowhere close to the theoretical
minimum size.

I did some more searching, and came across a few other posts from people
asking the same thing, and for some reason, it seems like the response
is always something along the lines of your initial statement: PNG is
lossless, so there's nothing you can do. I'm not sure where this
widespread misunderstanding comes from though. As an example, let's take
two trivial run-length encoding style lossless compression algorithm.
Say your input is a string of 32 "A"s, followed by 10 "B"s:

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABBBBBBBBBB

In algorithm 1, the maximum run length is 8, so that string would
compress down to:
8A 8A 8A 8A 8B 2B

In algorithm 2, the maximum run length is 256, so that same string would
compress to:
32A 10B

Obviously, algorithm 2 has a higher compression level than algorithm 1,
yet both are lossless.

Now that's a contrived example, but it's very similar to what real-world
lossless compression algorithms do, including the one that PNG uses.
Higher compression levels translate into using larger buffers to look
ahead further and find repeated strings, along with more memory usage
and CPU usage. If you've used archiver programs like WinZip or 7-Zip,
you'll see they let you control the compression level... there's often a
"fast" setting that doesn't compress as well, but is really fast,
"normal", and a "best" setting that is slow, but gives smaller file sizes.

And to get back to PNG, see the description of libpng's
png_set_compression_level() at
http://www.libpng.org/pub/png/libpng-1.2.5-manual.html#section-5.8

That is exactly what I'm looking for... sure, I could just _use_ libpng,
but it'd be a shame to have to go to an unmanaged DLL when MS supplies
as part of .NET something that almost does what I need.

Also see http://en.wikipedia.org/wiki/Pngcrush and
http://en.wikipedia.org/wiki/OptiPNG , two utilities which take a PNG
file and losslessly recompress it to a smaller size, if possible.

Anyways, hope this explains a bit more about PNG compression, and how
those of us who are looking for a way to adjust the compression level
aren't asking for something that doesn't make sense.

However, I do get the impression that it doesn't exist in
System.Drawing.Imaging :( At least
http://msdn2.microsoft.com/en-us/library/ms533846(VS.85).aspx seems to
say that the PNG encoder has no configurable parameters. So perhaps I'll
have to use libpng after all.