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
When a Fluent node writes into a channel, it expects the tuple to be sent to another Fluent node and placed in its channel. In order to know which channel the tuple should be placed in, each channel has to have a unique name. If all Fluent programs are running the same program, this can be checked at runtime when the program starts. If different Fluent programs are all communicating, then enforcing global uniqueness is probably impossible, but we can do something to make it harder to mess up.
This may be something we metaprogram in Fluent. That is, we write a Fluent program deployer that deploys Fluent programs. deployed checks invariants like unique naming, etc.
Alternatively, deployer could assign namespaces (name prefixes) for each program to be deployed, and install shim operators on the program's output and input that prepend and strip (respectively) the namespace to the Collections.
The text was updated successfully, but these errors were encountered:
When a Fluent node writes into a channel, it expects the tuple to be sent to another Fluent node and placed in its channel. In order to know which channel the tuple should be placed in, each channel has to have a unique name. If all Fluent programs are running the same program, this can be checked at runtime when the program starts. If different Fluent programs are all communicating, then enforcing global uniqueness is probably impossible, but we can do something to make it harder to mess up.
See more discussion in #3 from @jhellerstein :
The text was updated successfully, but these errors were encountered: