On 9/6/06, Anne van Kesteren <[EMAIL PROTECTED]> wrote:
On Wed, 06 Sep 2006 17:59:40 +0200, Mark Baker <[EMAIL PROTECTED]> wrote: > My reasoning is simply that it's advantageous to separate document > data and metadata (i.e. to not sniff) for reasons of security and > efficiency. I don't really see what you mean with the efficiency argument
I mean that I want to be able to hand off the incoming data stream to the correct processor at the earliest possible time to minimize latency, and since the media type arrives before the root namespace - and in plain text form (not encrypted or compressed) - it's more efficient to do so using its value.
and the security argument applies nonetheless given that you also want to support it for arbitrary XML media types.
I'm not sure what security issue you're referring to, but I'm referring to the kind that results from sniffing where documents can be crafted which can masquerade for other formats, bypassing firewall policies. There's also the issue of placing unnecessary constraints on XML language designers. The namespace of the root element isn't special, and I should be allowed to design an XML format which has a root element with any namespace. Like RDF/XML or XSLT, as mentioned before. Mark.
