- From: John J. Barton <John_Barton@hpl.hp.com>
- Date: Tue, 11 Mar 2003 10:22:33 -0800
- To: <eugene@datapower.com>, <xml-dist-app@w3.org>
At 08:21 AM 3/11/2003 -0500, Eugene Kuznetsov wrote: > > All we'd need is a "binary" rather than "base64binary" that did not > > apply base64 to the data. The base64 bit is only useful for email. > >Well, it's not as simple as just no encoding -- you still have to deal >with the delimeter problem, right? If your binary data contains >"</foo>", that would be a problem. What you need is an unparsed entity >of sorts for binary data. Amy's XML scheme extension for MIME content prefixes the non-XML content with a content-length. No delimiters needed, hence no delimiter problem. > > > > to outweigh the > > interoperability / backwards incompatibility costs, preferably with > > realistic use cases as actual performance/bandwidth data. > >Mike, where/to-whom would one submit these? Just the business of >optimizing handling of arbitrary-length element and attribute names can >keep someone quite busy for a while, multi-megabyte base64 sections will >not be a good thing. > >Interestingly, all the discussion focus here is on the on-the-wire >encoding issues -- doesn't seem that anyone feels strongly that binary >data should be passed by reference in some external envelope? On the contrary, I still believe that a layered message stack for mixing XML and non-XML data is a better design: Application e.g Control/Data XML/MIME-types Packaging e.g. SwA Transfer HTTP Transport TCP/IP (BTW RFC1341 makes a pretty good case for this design). This model leads to XML referencing entities in the package. Such a model separates the tool set for packaging from the tool set for application data, allowing each to be optimized for their task and ignorant of the detail of the other one. Tool-set developers, of course, prefer their tool to be used for the entire stack. In some recent mail we have been exploring alternatives that embed the binary in the XML, eliminating the packaging layer. I still believe that this is a technically inferior solution for large systems, but if its the only one that we can agree on I'd like to see if we can minimize the damage ;-) ______________________________________________________ John J. Barton email: John_Barton@hpl.hp.com http://www.hpl.hp.com/personal/John_Barton/index.htm MS 1U-17 Hewlett-Packard Labs 1501 Page Mill Road phone: (650)-236-2888 Palo Alto CA 94304-1126 FAX: (650)-857-5100
Received on Tuesday, 11 March 2003 13:22:37 UTC