Lists Home |
Date Index |
Bill de hÓra wrote:
> Robin Berjon wrote:
>> tend to be pretty confident I can make sense of it on both sides.
>> There's no reason a well defined binary infoset couldn't make anyone
> Yet somehow we haven't managed to do that yet. All the stuff that
> interoperates is textual.
That's why I feel we have to address the needs of people that send binary
infosets around. I think it would be a great loss if binary infosets keep
spreading while being only locally interoperable (as is currently the case
within MPEG-7, TV Anytime, and others).
> I wonder if the problem is not so much binary v XML, but the that XML
> trees are awkward enough datastructures to stream efficiently and
> require too much buffering. Even if you can pass it around binary
> Infosets, processors still have the root ndoe/element item to contend
That's a solved problem: at least one binary infoset format (BiM) is streamable
in any order, and in fragments. Entire subtrees can be skipped immediately if
you don't need them. In fact the skippability is what makes the format resilient
to change. That sort of feature is used a lot in broadcast environments where 1)
if you update your schema you can't be forced to update a few million set-top
boxes and 2) given that the data stream is permanent and unidirectional, you
need to be smart as to how you send your data. What happens is that the binary
infoset is constantly broadcasted (using carousels of data) and a TV set or
set-top box doesn't have to wait until it's received the entire document (or
seen the root) to start displaying its content. In fact, some more frequently
changing data (eg the score in a tennis game) is sent more often than the less
frequently updated bits (eg the player names) so as to obtain better reactivity.
Robin Berjon <firstname.lastname@example.org>
Research Engineer, Expway
7FC0 6F5F D864 EFB8 08CE 8E74 58E6 D5DB 4889 2488