Hi Anastasia, Is the concern that deserialization would have no guarantee as to the order > it appears? >
No, it's simply that when reading the logs screeds of unformatted JSON can be hard to read, especially when more important fields come later in an object (or after an array). This is also a problem with the existing format too, of course. And there's now the option to put it through `jq` to indent it. So I don't think it's worth obsessing over. > I understand the performance concern. However, this would only be logged > when the trace level logging is turned on, so this logging isn't done > otherwise. Plus, the existing text-based logging is already very expensive, > so it might not be that significant of a change. > Yeah, it's an implementation detail, so could always be optimised later when it's known to be a problem. Thanks again for the KIP, LGTM. Tom
