Bug 168467
Summary: | A/B testing | ||
---|---|---|---|
Product: | WebKit | Reporter: | JF Bastien <jfbastien> |
Component: | Web Template Framework | Assignee: | Nobody <webkit-unassigned> |
Status: | NEW | ||
Severity: | Normal | CC: | fpizlo, jfbastien |
Priority: | P2 | ||
Version: | WebKit Nightly Build | ||
Hardware: | Unspecified | ||
OS: | Unspecified | ||
Bug Depends on: | |||
Bug Blocks: | 168453 |
JF Bastien
I added a quick A/B test hack in bug #168453. We should do this better: factor it out, make it possible to A/B test other things, automatically create the random value, and prevent correlation between decisions.
I'd do the following: template<typename Int> Int abTest(uint32_t uniqueABTestValue, Int minimum, Int maximum);
This returns a value in [minimum, maximum] based on anonymous UUID, and mixes in the unique A/B test value so that each A/B test yields an independent result per user. This is important because users have their own UUID, but we don't want them to all be in the same side of each A/B test! Mixing in a unique A/B test value ensures that each A/B test skews a different way per UUID.
We need to use a simple hash function, so that from crash logs we can reconstruct which A/B test was set, and to which value.
Attachments | ||
---|---|---|
Add attachment proposed patch, testcase, etc. |