NewConstant panics if the given base is zero

[Deleplace]

It seems to me that a zero Backoff makes sense, from a user POV who would just want to enjoy a retry facility.

Can we make 0 an accepted value for NewConstant?

Of course it is not difficult for a user to just call NewConstant(time.Nanosecond), but for flexibility I'm not sure we should advertise "zero delay" as a forbidden parameter.

Zero delays may already occur when using the API, as discussed in #8.

[sethvargo]

Hey @Deleplace - I'm really hesitant to do that because it creates a hot loop. The default behavior (without middleware) is to retry indefinitely. Thus retry.NewConstant(0) with a non-terminating condition would completely hog all CPU cores until OOM.

[Deleplace]

Right, busy loops are not great.

The use cases I had in mind (network requests, DB requests, filesystem I/O) are so slow that they are self-throttled even without a delay between two attempts.

Now I realize that some use cases may indeed hog CPU/memory: whenever a resource is "not ready" and this information is in memory. E.g. "no network for now".

So, a zero backoff delay would be safe only when combined with flavors of WithMaxRetries, WithMaxDuration, WithJitter.

[sethvargo]

Even then, I think having a base like 10*time.Nanosecond is a good best practice.

[github-actions]

This issue is stale because it has been open for 14 days with no
activity. It will automatically close after 7 more days of inactivity.

[github-actions]

This issue has been automatically locked since there has not been any
recent activity after it was closed. Please open a new issue for
related bugs.