PDFs and encoding

Hi,
I'm looking for resources on the above.

I'm confused about the relationship between the encoding of sources [sgml], the encoding 
of the body text within the generated PDFs, and the impact/restrictions they have on the 
encoding of the metadata in the trailer part of the file.

I'd appreciate a pointer to some resources.

Particularly interested in unix cmd line tools that will add multibyte metadata to a PDF, 
after the PDF has been created.

Thanks,

~mm