bassosimone
commented
7 months ago With https://github.com/ooni/probe-cli/pull/1392 and all the other PRs referenced above, we are now able to produce the same top-level test keys for v0.4 and LTE as far as our QA suite is concerned. Hopefully and ideally this addresses the concern that switching to LTE is going to cause many differences with respect to v0.4.
I solved the issue by allowing LTE to choose different engines for data analysis. The previous code and engine is now called "orig" and is not the default. A new engine, called "classic", which is now the default, attempts to be very compatible with v0.4. As follow up work, we should figure out a way to mix these two engines, such that we get both compatibility with v0.4 and more advanced flags that characterize all forms of blocking.
In any case, we're now ready for releasing and doing https://github.com/ooni/probe/issues/2555.
We have a common test suite for webconnectivity v0.4 and v0.5 based on netemx. We should perform the following set of activities: (1) write more test cases and (2) make sure the differences between them are not regressions. I already know that there are differences that are regressions. We should try to act upon those differences in a short time frame (e.g., a sprint) and we should document as follow-up issues any other differences we could not fix. Once this issue is done, we can done proceed to A/B testing in the real world through https://github.com/ooni/probe/issues/2555.