On Wed, Aug 27, 2003 at 12:52:48PM +0200, Robin Berjon wrote:Yes, I'm aware of the issue, which is probably why I ended up on the program committee of this[0] Thing With A Horrible Name :)
[0]http://www.w3.org/2003/07/binary-xml-cfp.html
Pardon me for being thick here, but what possible gains are there for this in general-purpose use over just gzip / zlib ing the data ?
I tend to put speed, size, random access, dynamic update on my list of benefits. Other people have other views (often a subset of that). Sometimes gzip doesn't cut it, either because even with higher CPU you may have battery-usage problems (eg on cell-phones) or because you want much better compression (I've gotten x200 compression on some SOAP messages where gzip got around x3 for intance).
Random access has all sorts of uses, for instance it's dead useful for an XML editor (it can change a document in the middle of it, instead of reading it all, changing it, and writing it all out). Dynamic update is very good for very low bandwidth situations (where very low can be what you'd call really low, or can be something much bigger but you have lots of data to send). It is also critical for unidirectional systems such as digital TV broadcasts (putting SVG on the telly).
There are lots of reasons for various people. The workshop has received many more papers that I (and I think anyone) expected. Ask me in private mail if you want my own position paper.
IME the parsing of a non-trivial binary format is a much nastier task than parsing text, especially structured text like XML.
XML is a fairly complex structured text format if you take everything into account. And the binary format doesn't have to be non-trivial. Given a simplified schema such as RNG after simplification (as was discussed here earlier) the format can be fairly simple (just flag presence/absence of possible grammatical occurence).
Vendors are generally pretty shit at providing even a decent interoperable
*text*-based protocol (viz variously and gloriously broken HTTP/1.1 implementations (c) All Major Vendors). HTTP is, I would submit, hardly a
particularly difficult or fragile protocol to get right, yet vast numbers of
software companies seem incapable of getting it right or handling its edge cases sanely.
I agree.
On the binary format front, would we like to consider the resounding triumph of clarity and interoperability represented by CORBA or RMI-IIOP or any of the other nasties out there?
Those are domain specific.
I can't see that reinventing the wheel with a bunch of domain-specific binary formats is anything other than a retrograde step.
We're not gunning for domain-specific. We're gunning for generic. It won't happen if it's not simple.
PS: Ben, I've answered your other private mail but your SMTP server blacklists the entirety of wanadoo.fr, which honestly is quite a bad idea :)
-- Robin Berjon <[EMAIL PROTECTED]> Research Engineer, Expway http://expway.fr/ 7FC0 6F5F D864 EFB8 08CE 8E74 58E6 D5DB 4889 2488
