Re: [binaryXML-30] Binary XML problem statement.

On Wed, 2003-02-19 at 14:20, Noah Mendelsohn/Cambridge/IBM wrote:
> Mike:  I think you have written a wonderful and balanced summary.   One 
> comment (as much to Chris as to you):  Chris' summary says [1]:
> 
> "The primary reason that people give for using Binary XML (binary 
> representations of an XML Infoset) is size efficiency - both in network 
> transmission and in storage on the receiving device."
> 
> Actually, in the Web Services area, I've heard at least as much interest 
> attributed to (perceived) improvements in 
> serialization/parsing/deserialization time.  Remember, some of these web 
> services deployments are attempting to displace systems in which you read 
> in the C structure off the wire, set a pointer to it, and go (assuming 
> byte order, security, and versioning aren't issues.) 
> 
> As a somewhat extreme point of comparison, you wouldn't for most purposes 
> want to represent an IP packet in a format with variable offsets, allowed 
> whitespace, etc.  IP packets aren't the subject of discussion here, but 
> many of the existing protocols have similar characteristics.
> 
> I'm not particularly an advocate for binary XML at this point, but I agree 
> with Mike that we need an orderly cost/benefit analysis, and pathlength is 
> at least as much a concern as space in certain environments.  Thank you.

While not quite the proto-typical web services area, there are several
IETF protocols that would like to use XML that also have these same
characteristics. In several cases we have requirements that closely
match DNS's 'network footprint' and server side scalability strengths.
Thus for many of those just applying 'gzip' between the serializations
defeats the purpose. Instead we're after something like you describe
above: we would like to receive some small very small XML, pass a
pointer to the XML parser and 'just go'.

-MM

Received on Wednesday, 19 February 2003 15:10:10 UTC