Re: [xml-dev] First Working Draft of the XML Binary Characterization Use Cases

Hi Tony,

Anthony B. Coates wrote:
> Good work to everyone involved.

Thank you!

> I think the FIXML use case can be  
> expanded to cover much of the financial community.  It is fair to say 
> that  lack of network bandwidth has been the major impediment to the use 
> of XML  in finance.  There are still a lot of 64K circuits in use, and 
> when there  is other data to fit into that bandwidth as well, you can't 
> get away with  sending XML over the wire as ASCII.

Several of the use cases in the document could be (and will be) 
expanded. I think it's fair to say that we are covering a fair amount of 
ground but it's true as well that we can't possibly have representatives 
from every single industry that's using XML, or even just those that are 
trying to use XML but having problems.

Therefore, if you (or more generally LMS) would like to expand our use 
cases, we are all ears. From what you say I gather that the main problem 
you are facing is size, and that you can't address is with conventional 
generic compression because of issues with transparency of the encoding.

The size problem is one that is found in many of our use cases, and most 
of them list reasons why generic compression doesn't fly. What's 
different in your case is that while the others generally have issues 
with the fact that compression adds extra cycle, your issue with 
transparency doesn't appear in the others.

I must say that I am slightly surprised with it, and would like to hear 
more on the topic. First, a lot of compression can happen at the 
transport level, where it will be transparent to the applications 
anyway. I suppose you have looked at that and have reasons as to why and 
how it is insufficient. Second, we have heard the statement several 
times that implementing the type of transparency you are describing 
isn't difficult -- in fact, some XML parsers such as libxml supporting 
reading directly and automatically from a gzip stream, and all SVG 
implementations are required to do the same, even in situations where no 
metadata (such as a Content-Encoding header) is available. Again, I am 
sure there are reasons why the financial community is having such 
problems, and in order to take those issues into account, we would very 
much like to have a better grasp for them.

Thank you for your feedback!

-- 
Robin Berjon

Received on Monday, 2 August 2004 06:36:41 UTC