Re: [xml-dev] Re: [ubl-dev] Top 10 uses of XML in 2007

[Elliotte Harold <elharo@metalab.unc.edu>]

Stephen Green wrote:

> Isn't this all just too obvious to debate?
> 

No, it isn't. A lot of people think it is, but it's not. In fact, there 
just aren't the measurements to prove the advantages most people claim 
in general purpose scenarios.

The binary folks have mostly forgotten (or never knew) the first and 
second rules of optimization. The first rule is don't do it.

The second  rule (and this one's for experts only) is don't do it 
without very careful measurements of exactly what it is you're trying to 
optimize and exactly what it is that's causing those operations to take 
longer.

The bottom line is that most of us rarely if ever need anything smaller 
or faster than XML. I think the proof of this is how much slower and 
larger a lot of binary file formats such as Microsoft Office's are in 
practice. If anyone really needed these to be smaller, they would have 
complained about that, irrespective of whether it's binary or not.

Any message that starts off with the claim that we must make XML smaller 
or faster by changing it to binary is suspect. That is not where 
informed optimization starts. Informed optimization starts with 
measurements of systems and demonstrations that the existing performance 
  is inadequate for the use case. This is followed by profiling to 
demonstrate where the time (or size, or battery life) is actually going.

Only if those two steps demonstrate that XML is actually the problem can 
you then reasonably consider improving or replacing it. Anybody who 
starts off with a claim that we have to go to binary, or for that matter 
starts off with any claim related to performance without those two 
prerequisites, is wasting our time. If there even is a problem (more 
often than not there isn't) it's almost never what they think it is.

The only folks I've heard make any reasonably informed claims about XML 
issues that might suggest going to a different format are in the mobile 
community, and their major problem is battery life, not size.

-- 
Elliotte Rusty Harold  elharo@metalab.unc.edu
Java I/O 2nd Edition Just Published!
http://www.cafeaulait.org/books/javaio2/
http://www.amazon.com/exec/obidos/ISBN=0596527500/ref=nosim/cafeaulaitA/