[04:30:17] milimetric: yeah [04:30:28] yeah :) [04:30:28] ? [04:30:35] you should probably not be able to run an hourly time series over more than like, a week. [04:30:37] maybe a month [04:30:40] heh [04:30:42] right, not a year [04:30:54] if we wanted to make a logical strict rule [04:30:56] it might be [04:31:11] you can't run a time series that's greater than the sum of its parts [04:31:30] no hourly series over more than a day, no day series over more than a month, no month series over more than a year [04:31:40] but that's maybe being too hard [04:31:55] well, maybe the rule should just be about how much data is generated [04:32:09] if you only have one user, it doesn't hurt anyone if you look at hourly over a month [04:32:14] I guess [04:32:33] but hm... then if you wanted to know interesting stuff about large cohorts... [04:32:57] Yeah, it does penalize people who have larger cohorts and let small cohort analysts be crazy [04:32:58] maybe your rule to the second degree - no hourly over a month, daily over a year [04:33:07] that makes sense [04:33:20] k, i like it, i'll suggest to list [04:33:21] thanks! [04:33:31] NP [04:34:19] oh! [04:34:33] I also wanted to thank you for the feedback during the quarterly [04:35:09] I'm trying to figure out how to sqoop data from some obscure wiki into hadoop so I can start building a decent cross-wiki warehouse for you guys [04:35:26] I'm more and more convinced that the key to getting good work done here is really just being bold [04:43:40] Yes [04:43:43] and you're welcome [04:44:09] Ori is right when he says that at WMF resources/energy coalesce around successful projects. [04:44:36] So if you can prove your pudding, so to speak, good things will happen [05:08:05] cool, it works, I'm thinking this was never very hard. And the sqoop docs are written for dummies like me. <3