Jonathan Lange

Well, there's a `serialize` method, which would address it too, and with less code than #95. Although it would still need to be documented somewhere other than the source code....

Sounds pretty sensible. Would suggest making the filtering abstraction completely unaware of ring buffer and the like, and providing an implementation that provides the ring buffer stuff. Not 100% clear...

Agree that a new, entirely-eliot log is the better way. I also think that if we do that then the tests themselves should be treated as eliot actions. I've made...

Temporary workaround: https://github.com/jml/trial-eliot/blob/master/bin/from-trial-log Run that on a `_trial_temp/test.log` and it will do its best to extract the eliot messages and convert them to JSON.

Thanks for the bug report. This is a good point. Falling back to plain grep is probably a sensible default.

The problem is that the code snippet in the diff is fundamentally ambiguous. Pygments is choosing to interpret the docstring as _beginning_ on the third line, rather than ending. This...

Why is this a problem? Is it that the function is doing unnecessary extra work? On Wed, 21 Feb 2018 at 13:47 Bryan Boreham wrote: > I.e. this line: >...

Don't know. I'm content for it to be closed.

It seems reasonable, although I'd personally prefer something like an HTTP 204 on successful bind. I can imagine alternate server implementations where `bind` might fail (invalid app id; invalid side;...

Thinking about this a little. It seems a little inconsistent that `released` and `closed` are returned but that there's no `bound` (or `opened` for that matter). The inconsistency doesn't make...