[issue19183] PEP 456 Secure and interchangeable hash algorithm

[Nick Coghlan]

Nick Coghlan added the comment:

I reviewed the latest PEP text at http://www.python.org/dev/peps/pep-0456/

I'm almost prepared to accept the current version of the implementation, but there's one technical decision to be clarified and a few placeholders in the PEP that need to be cleaned up prior to formal acceptance:

* The rationale for turning off the small string optimisation by default rather than setting the cutoff to 7 bytes isn't at all clear to me. A consistent 3-5% speed difference on the benchmark suite isn't trivial, and if we have the small string optimization off by default, why aren't we just deleting that code instead?

* A link to the benchmark suite at http://hg.python.org/benchmarks should be included at the appropriate places in the PEP

* The "Further things to consider" section needs to be moved to a paragraph under "Discussion" describing the current implementation (i.e. the hash equivalence is tolerated for simplicity and consistency)

* The "TBD" in the performance section needs to go. Reference should be made to the numbers in the small string optimisation section.

* The performance numbers need to be clear on what version of the feature branch was used to obtain them (preferably the one you plan to commit!).

----------

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue19183>
_______________________________________