xitep

ah ... i think i wasn't clear in my previous comment and misled you. sorry for that. my point was this: - you suggested to generify `KafkaConnection::new` over the `ToSocketAddrs`...

yeah, seems like that would do the trick! :) :+1: i see two disadvantages though: a) giving up on hostnames will make the log output less friendlier to use (mine...

unfortunately, no. not that i'm aware of. feel free to tackle this one. my focus is currently rather on #63.

true enough; looks like the special case mentioned in #41

@je-ik i think your case suggests that the validation mentioned in this ticket is to be done only _after_ the expansion of to operators to their basic "atoms." (or rather...

@je-ik sorry for the delay. i think i got your point. yes, it makes sense, absolutely. i just don't have any particular idea about how we could differentiate the two...

Made a start [here](https://github.com/seznam/euphoria/wiki/Features). Feel free to contribute.

Alternatively, another suggestion has been made by @je-ik : Client code as part of a flow: ```java Dataset input = ...; Dataset transformed = Transform .named("mytransform") .of(input).return(Y.class).output(); ``` Further register...

while we achieved a lot, i fear the pure overhead on flink is considerably greater than we thought so far. i just happened to run the streaming [benchmarks](https://github.com/seznam/euphoria/tree/master/benchmarks) on text...

regarding my last comment: as usually the problem was sitting on my chair ;) the reported "3 minutes" runtime was observed while flink ran with a heap storage backend. after...