You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Boshen has mentioned telemetry a few times. I also think it would be really helpful to have some numbers on what kinds of projects people are using Oxc on e.g.:
How many files?
Proportion of those files which are JS/JSX/TS/TSX?
Range of size of files (measured both in KB, and in number of AST nodes, scopes, references etc)?
oxc-project/oxc#4328 is a good example of where optimizing for one size of file degrades perf on another size.
I was asserting on that PR that, of our benchmarks, RadixUIAdoptionSection.jsx is the most representative of real-world application code. I think that's probably true, but actually I have little evidence. Let's gather some!
How to gather it
At present, oxclint has the widest adoption. Can we build telemetry into it?
Ideally it would gather info about not just the files which are linted (application code), but also about the mass of code in node_modules which Oxc will also be acting on via e.g. Rolldown.
Probably adding the actual measures is fairly simple, but:
Where do we store the data that's uploaded?
How do we give user option to opt out? (config file? command line switch?)
How do we communicate the reason for telemetry? We'll want to assuage potential concerns that it's "spying", explain how data is gathered and used, and encourage users not to disable it.
Does GDPR and similar legislation come into this? (hopefully not)
Leave it to Rolldown?
Or... if Rolldown plans to add telemetry, do we leave it to them to do, and then use the data they gather?
The text was updated successfully, but these errors were encountered:
Why we need telemetry
Boshen has mentioned telemetry a few times. I also think it would be really helpful to have some numbers on what kinds of projects people are using Oxc on e.g.:
oxc-project/oxc#4328 is a good example of where optimizing for one size of file degrades perf on another size.
I was asserting on that PR that, of our benchmarks,
RadixUIAdoptionSection.jsx
is the most representative of real-world application code. I think that's probably true, but actually I have little evidence. Let's gather some!How to gather it
At present, oxclint has the widest adoption. Can we build telemetry into it?
Ideally it would gather info about not just the files which are linted (application code), but also about the mass of code in
node_modules
which Oxc will also be acting on via e.g. Rolldown.Probably adding the actual measures is fairly simple, but:
Leave it to Rolldown?
Or... if Rolldown plans to add telemetry, do we leave it to them to do, and then use the data they gather?
The text was updated successfully, but these errors were encountered: