Thanks for the PR!

I understand the problem as stated in the PR, but am a bit hesitant about making the change at this layer -- I think it makes sense for the core data structure's serialization and deserialization be more well-defined/stringent.

It feels like trimming would be more well-suited at the layer that reads the file from disk -- is that controlled in ECV or by client libraries? I would much prefer a solution that trimmed the input that is read from the file.