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
At this time, backoff supports the usecase well, where the code calling RetryNotify(b, n) is either directly related to the notification function passed or at least got it passed in explicitly along with a particular backoff.
When RetryNotify(...) is deeper in a layered architecture then a higher layer or the upmost layer might be interested in notifications instead, as it is the ultimate consumer as well as the originator of the backoff. In order to avoid having all intermediate layers having to be made aware of notification callbacks, would it be possible to attach notification callbacks to the backoff objects themselves?
This would allow the source of backoff information to also be able to get ultimately informed about any retries without much hassle.
Some use cases might be service logging by the ultimate service as to not having to instrument lower layer modules with logging and specific loggers (a nightmare in itself), as well as easy uni test instrumentation.
At this time, backoff supports the usecase well, where the code calling
RetryNotify(b, n)
is either directly related to the notification function passed or at least got it passed in explicitly along with a particular backoff.When
RetryNotify(...)
is deeper in a layered architecture then a higher layer or the upmost layer might be interested in notifications instead, as it is the ultimate consumer as well as the originator of the backoff. In order to avoid having all intermediate layers having to be made aware of notification callbacks, would it be possible to attach notification callbacks to the backoff objects themselves?This would allow the source of backoff information to also be able to get ultimately informed about any retries without much hassle.
Some use cases might be service logging by the ultimate service as to not having to instrument lower layer modules with logging and specific loggers (a nightmare in itself), as well as easy uni test instrumentation.
This might be easier to achieve in terms of API with the "builder pattern" issue #106 in place.
The text was updated successfully, but these errors were encountered: