Zlib-compatible compression streams?

Question

Are System.IO.Compression.GZipStream or System.IO.Compression.Deflate compatible with zlib compression?

Answer 1

I've used GZipStream to compress the output from the .NET XmlSerializer and it has worked perfectly fine to decompress the result with gunzip (in cygwin), winzip and another GZipStream.

For reference, here's what I did in code:

FileStream fs = new FileStream(filename, FileMode.Create, FileAccess.Write);
using (GZipStream gzStream = new GZipStream(fs, CompressionMode.Compress))
{
  XmlSerializer serializer = new XmlSerializer(typeof(MyDataType));
  serializer.Serialize(gzStream, myData);
}

Then, to decompress in c#

FileStream fs = new FileStream(filename, FileMode.Open, FileAccess.Read);
using (Stream input = new GZipStream(fs, CompressionMode.Decompress))
{
   XmlSerializer serializer = new XmlSerializer(typeof(MyDataType));
   myData = (MyDataType) serializer.Deserialize(input);
}

Using the 'file' utility in cygwin reveals that there is indeed a difference between the same file compressed with GZipStream and with GNU GZip (probably header information as others has stated in this thread). This difference, however, seems to not matter in practice.

Answer 2

I ran into this issue with Git objects. In that particular case, they store the objects as deflated blobs with a Zlib header, which is documented in RFC 1950. You can make a compatible blob by making a file that contains:

• Two header bytes (CMF and FLG from RFC 1950) with the values 0x78 0x01
  • CM = 8 = deflate
  • CINFO = 7 = 32Kb window
  • FCHECK = 1 = checksum bits for this header
• The output of the C# DeflateStream
• An Adler32 checksum of the input data to the DeflateStream, big-endian format (MSB first)

I made my own Adler implementation

public class Adler32Computer
{
    private int a = 1;
    private int b = 0;

    public int Checksum
    {
        get
        {
            return ((b * 65536) + a);
        }
    }

    private static readonly int Modulus = 65521;

    public void Update(byte[] data, int offset, int length)
    {
        for (int counter = 0; counter < length; ++counter)
        {
            a = (a + (data[offset + counter])) % Modulus;
            b = (b + a) % Modulus;
        }
    }
}

And that was pretty much it.