Yes, but both BZ2 and GZIP are very good at working with repetitive
strings. I haven't looked much at the EBML, but my impression is that the
XML file will be much more compressable than the EBML, and if the XML is
larger than the EBML uncompressed is completely irrelevant.

We have here a lot of speculations (on both sides). Perhaps we should make some real tests and take our own sharelist (we should be enough to have a reasonable results), converts it into XML and EBML and then compress it with BZIP2 and GZIP. Then, we will have at least a little base of discussion.

XML has the extra advantage of being a standard with several
implementations on different platforms and in different languages which has
to count for something. There are plenty BZ2 and GZIP implementations
available aswell.

You are fully right, EBML is probably to recent to have a port everywhere.

Eric