-
Notifications
You must be signed in to change notification settings - Fork 22
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add willReadFrequently: true to Images test #1
base: main
Are you sure you want to change the base?
Conversation
Do we have any data to back up the claim that this is what authors are doing? |
We expect authors to use this path in the wild so we should prefer measuring it.
LGTM |
This should be a separate subtest, I think. I don't think we should modify the existing subtest - it intentionally doesn't test the willReadFrequently path |
Is the current non- i.e. do you have evidence that it's commonly hit and improving it makes a difference for users? |
Well,
If the request is for us to gather telemetry for all the times |
The request would be for telemetry for frequent reading without |
@junov, in https://bugs.webkit.org/show_bug.cgi?id=235196#c6 you mention adding instrumentation to measure feature adoption. Do you have the results of that instrumentation that you can share? |
(FYI: Us gathering telemetry will take about a year) |
I also don't understand the resistance to making a separate test for this. What's the downside? |
I think the resistance is that we don't really want to incentivize spending engineering on performance optimizations for a path that we actively discourage authors from using. We likely want to track the willReadFrequently:false testcase individually in our own regression testing, but I'd prefer not to encourage engineering investment in something that, should an author wish to, is loudly warned about and easily fixed. I think we're not totally against having tests for both, but that I feel like we'd prefer to weight them as something like 30:1 for willReadFrequently:true vs false. In practice, this is akin to not having the willReadFrequently:false/default case. |
We expect authors to use this path in the wild so we should prefer measuring it.