Pathological DFDL Parser behavior

[mbeckerle]

DFDL is (we believe) a finite, non-turing complete language. But, DFDL schemas have been created that can construct an infoset that for input data of size N, create a super-exponentially larger infoset than N.

See https://github.com/OpenDFDL/examples/tree/master/superexp where the README.md file explains the issue.

This behavior is unintuitive. People expect that for data of size N and a schema of size S that the infoset can at worst be O(NS).

---

Bradd K mentioned that customers use the transformation capabilities of DFDL to create and populate structures that are different from the input format, and in unparsing to populate the output from structures that are different.

So any solution to the pathological DFDL behavior problem can't prevent these kinds of transformations since users are depending on them.

However, there is nothing that requires creation of non-linearly larger structures.