Sesame's SPARQL client currently sends multiple Accept headers with its query requests, to indicate which response formats it can process. For example:
While this is technically correct (see section 4 in RFC2616 for details), it appears many SPARQL endpoint implementations have trouble processing request headers in this form. E.g. in the above example, they will choose to use text/csv (even if they do support application/sparql-results+xml).
To facilate better interoperability, we should instead compose a single accept header, with the supported formats as comma-separated values: