Here's some info about data URIs that may be useful in your research:
 
My data URI composer at <http://shadow2531.com/js/create_data_uris.html> is based off of stuff I learned about data URIs when writing some patches for <https://bugzilla.mozilla.org/show_bug.cgi?id=548516>. Firefox's data URI parsing code is very simple to the point that the metadata in the URI has to be perfect or Firefox will break.
 
The bug has a lot of comments. But, there are multiple ways of parsing and handling data URIs and browsers don't exactly do the same. Opera and Safari for example seem to parse data URIs like a mime header named "data" (after percent-decoding the first part of the data URI). One of my proposed patches makes Firefox work like that too. Firefox right now just does very basic (and brittle) string processing.
 
In the public-uri list, I have a few threads with stuff like this.
 
"data URIs - filename and content-disposition" (<http://lists.w3.org/Archives/Public/uri/2010Mar/0004.html>)
"Data URI handlng (was: Re: data URIs - filename and content-disposition)" (<http://lists.w3.org/Archives/Public/uri/2012Mar/0000.html>)
 
There's also <http://shadow2531.com/opera/testcases/datauri/data_uri_rules.html> that I came up with a while ago for parsing data URIs. The idea of that was to make Firefox's code less brittle and add support for filename and content-disposition threads. It's an old draft, so don't read too much into it. But, it shows things to think about when parsing data URIs.