This a nasty fact that XML is still perceived as slow and fat.
This is globally non accurate in many ways.
In many area, we can reach processing time of XML as high as binary data carefully crafted data. But to achieve this goal, we need to come up with a stack of tools that all meet those requirements.
One of the goal of this community group is exactly to give some feedback about processing XML efficiently and showing to everyone that XML is not that slow and fat, and we can come up with reasonable and careful choice to very efficient processing time