RE: mnot-03, Infoset & syntax in sect. 2

["Bob Wyman" <bob@xxxxxxxx>]

David Orchard wrote:
> Atom may have a subset of the overall requirements that 
> allows it to have a particular encoding format that is 
> more efficient than any potential 80/20 solution that 
> covers all vocabularies.
	It is certainly the case that one could produce a binary
encoding that is so optimized to the requirements of Atom that it
would do a much better job (however "better" is defined) at encoding
Atom data than a general solution would do. This is especially the
case if the general solution is based on the Infoset. The same can be
said for any of the dozens or hundreds of XML based standards that we
have today. However, producing such a customized format for Atom (and
the others) is precisely the sort of thing that is going to result in
interop problems in the future. When each standard has its own
encoding formats, babel results. 
	Given this, I argue that Atom and others should *not* define a
custom, non-XML, encoding unless there is some exceptionally
compelling reason to do so. Personally, I don't see such a compelling
reason in the Atom case. Similarly, I argue that Atom should do
nothing to prevent the disciplined use of other encodings -- if such
standard encodings appear and become standardized in arenas more
global than the specific Atom context. 
	My objection in this thread was to the suggestion that "any
suggestion" of binary encodings would be fought "to the death." All I
am asking is that people keep their minds open and do nothing that
would prohibit or make difficult the use of a properly defined general
binary alternative if such a thing should become accepted.

> For example, the ratio of tags to content in a 
> sampling of feeds could be analysized and a particular
> solution based upon that be used.
	The problem with such an approach is, of course, that we can't
anticipate today what the usage for Atom will be in the future. Thus,
assumptions which are made based on any samples that are taken will be
speculative at best. For instance, we're already seeing debate on the
issue of whether an RSS feed should contain the entire content of an
entry or a summary of it. Feeds which contain an entire entry are
going to be drastically different than feeds that don't. Similarly, as
future MIME types are accepted, as people tend to use images more or
less in the future, etc. we'll see that the "average" characteristics
of an Atom feed change over time. It would be dangerous to make
long-term assumptions based on anecdotal evidence derived from
analysis of the Atom feeds we have available today.

		bob wyman