Atom and Binary XML (was RE: Re[2]: mnot-03, Infoset & syntax in sect. 2)

["Danny Ayers" <danny666@virgilio.it>]

(Bob Wyman suggests [1] Infoset-based binary serialization of Atom
syndication format, providing a schema in ASN.1 [2])

[Dave Pawson]
> Surely one of the standards bodies is the right place to resolve a
> binary format?  Perhaps in preference to this forum? Certainly take
> the use case you quote to such a forum.

Back in permathreadland...
I'm glad the issue came up, but I don't think it can be carried much further
in the context of Atom. Whatever the merits of binary XML, I suspect that
Atom wouldn't make a very good vehicle for its adoption. There is a
historical precedent:

It's apparent now that the RSS 1.0 format as RDF was ahead of its time. The
result is that there is now a *lot* of RSS/RDF data being syndicated
(perhaps 1/3 of feeds) yet very few of the current RSS tools actually use
RDF, treating RSS 1.0 merely as just another XML language. Sure, use of RDF
in aggregators is likely to increase - RDF adoption is growing fast outside
of syndication and RDF tools can consume and process RSS 1.0 directly. But
I'm not sure the rate of adoption of RDF *within* syndication would have
been significantly different had it come from the outside in. With a little
pre-processing (e.g. XSLT), RDF tools can also consume and process vanilla
XML RSS 2.0 data.

The bottom line is that despite all it's benefits and (gorblimey!)
appropriateness for Atom, which is in effect a resource description
language, the Resource Description Framework interchange format (RDF/XML)
isn't being used, and perhaps more significantly the mapping to the RDF
model is considered secondary. But RDF doesn't need Atom, Atom doesn't need
RDF. As long as Atom isn't designed as a stovepipe application language
(i.e. newsreaders-only, like RSS 2.0) then interop at a sophisticated level
will be possible. In the unlikely event that the Atom format became a
narrow, unextensible language, then standard tools will still allow baseline
interop with RDF.

So assuming that binary XML would in practice offer the benefits Bob and
others have described, I'd suggest that until it is adopted elsewhere there
is still nothing to really justify binary Atom, for either the syndication
community or the binary advocates. If binary XML is as good as advertised,
it doesn't need Atom and Atom doesn't need binary XML. Standard (maybe
ASN.1) tools can still interop. It's not hard to imagine the
uber-aggregators and mega-aggregators communicating with each other in this
way, sharing as much as possible of the bandwidth load of spidering.

My own opinion is that the benefits of binary XML over text XML are only
significant in edge cases, and in most of those a custom, non-Infoset-based
serialization would almost certainly be better. Having a completely
different XML serialization adopted widely could significantly screw up the
benefits a common format offers.

Another side question, assuming that binary Atom was available, how would
this work with the Atom API? The API is designed to be consistent through
all operations using RESTful techniques, including the transfer of feed data
from producer to consumer. Would binary Atom play nice with e.g. POSTing an
entry to a weblog? Dear permathread, just how RESTful is binary XML?

Cheers,
Danny.

[1] http://www.imc.org/atom-syntax/mail-archive/msg01838.html
[2] http://www.imc.org/atom-syntax/mail-archive/msg01849.html