Yes, the subject of DFDL infoset explosion interests me, and I took a look at
the personal github PR. I agree with the README's premise that the infoset's
size should be linearly related to the input data's size by O(sn). A runtime
check for forward progress on the problematic DFDL 1.0 feature
I have been concerned for a while that combinations of the features in DFDL
v1.0 can be used in ways that are just far too powerful for what is needed
for data format expression.
For example, if the input data to parse is n bits, how large should the
corresponding infoset be when the data is parse