On Wed, Mar 18, 2015 at 9:42 PM, Danny Horn dhorn@wikimedia.org wrote:
Brad: unfortunately, it's really hard to tell very much from a conversation with messages like "3: Post C: reply to Post A". You could do that with the old model, the new model or the perfect magic Nobel-Prize-winning discussion threading still to be discovered, and it would probably look like nonsense in all three.
I shouldn't have used both numbers and post-names, but once I realized that it was already a bit late and it won't let me edit those posts. Someone with appropriate permissions is free to go back and edit them all to remove the number prefix and let the alphabetical order of the post-names suffice to indicate the chronological order of the postings, if that would make it less confusing for you.
The point is the structure you're displaying doesn't make any sense, not that the content of my messages isn't anything that might make sense on its own. My "content" is explicitly simplified to illustrate the failings in the displayed structure. Structure should *facilitate* understanding, but in your demo I'd find that understanding the underlying structure of the conversation would be *despite* the broken display-structure.
Nor is the point that people can screw up wikitext talk pages in ways that are even more confusing. That's a given, but Flow is supposed to do better. Right now it's worse than a well-formatted wikitext talk page (which has the advantage that human users can *fix* the structure when a newbie breaks it).
Comparing http://flow-tests.wmflabs.org/wiki/Wikitext version of Topic:Sdrqdcffddyz0jeo to http://flow-tests.wmflabs.org/wiki/Wikitext_version_of_Topic:Sdrqdcffddyz0je..., I find it much easier in the latter to see what is a reply to what.
We've tried in our testing to pretend that we're having real conversations, so we could see whether there's any logical way to get to eight levels of nested threading. It's not easy to organize make-believe conversations, but if you want to start a thread, I'd be happy to fire up a few sockpuppets and pretend to talk about something with you.
No thanks. Pretend "real" conversations are ok for a first assessment at usability, but by nature they're likely to be vapid and unlikely to have the inter-post complexity of actual conversations on-wiki where people are concentrating on actually communicating rather than on forcing a conversation for the sake of testing.