[01:08:08] Amir1|afk: I need to look at the sql escaping a little more closely, will do so tonight [01:08:37] thanks legoktm [01:08:39] :) [01:08:42] really afk now [01:13:29] o/ [19:06:43] halfak: Do you consider it plausible that ORES might one day run in production and produce answers to model.reverted in under 150ms for the 95th percentile of edits? [20:25:45] hey Kinkle [20:25:47] Re. production, we're actively working on that. We have hardward and are configuring it. [20:26:13] Re. sub 150ms, I think that might be difficult, but I think it is possible. [20:26:47] One option that might work nicely is to use RCstream or Event bus as a means to maintain some caches. This would allow for faster data access. [20:27:10] However, computing a diff between revisions will always be some version of slow for many pages. [20:27:35] BUT, we'll likely get a good benefit from implementing the tokenizer and differ algorithms as a C library. [20:27:54] There've been some volunteers who picked that up, but I don't think any real progress has been made. [20:28:15] Why is sub 150ms an interesting threshold for you? [23:39:05] * halfak is having fun at conference, but is really looking forward to getting back to work.