> So what are the fundamental problems with using a text stream? Is it
> just hide-bound belief in binary streams or are there solid valid
> arguments to be made?

Delimiter. In XML you have to replace <,>,' and " with their escaped
equivalent. So we will have to invent something new. What does pure
text brings over the same binary representation unless you're going to
use RFC 2549 ("IP over Avian Carriers with Quality of Service") as a
way to connect to Firebird. :)

Another question to Jim:

What part in the current data encoding limits introduction of new
types? Do you plan to eliminate XSQLDA and XSQLVAR? If no, where's the
limitation?

According to your requirements current wire protocol does not meet the
"transmitted across machine architectures without reformatting or
conversion" in case of integers - wire contains big endian numbers, we
have to convert them to little endian numbers.

Personally I believe that density is not too important and I'm not
sure that your scheme is better compared to simple zip on top of
current protocol. In my previous company two guys invented a
compression scheme for vector data to fit the GIS system on PDA; it
included quite complex bit shifting and so on. They were very proud of
it until new CIO came into the house and compressed their data files
with zip with 25% smaller size.

The only argument I would buy is the extensibility. So why the current
approach is not acceptable?

Roman