=== Ng_ is now known as Ng [19:21] * mpt compares the layout of and [19:22] I suppose vs. is a fairer comparison [19:22] and they're only slightly different [19:37] newz2000, why do you use TM instead of ™? [19:37] I don't think I did that [19:37] mpt: where is it? [19:39] newz2000, http://www.ubuntu.com/getubuntu/download [19:39] Standard personal computer (x86 architecture, PentiumTM, CeleronTM, AthlonTM, SempronTM) [19:39] oh, I copy and pasted that from somewhere [19:39] but I probably didn't know about 8482 [19:39] too many entities these days [19:40] ™ is easier to remember :-) [19:40] are you looking over that email I sent re the download page? [19:40] yes, that's why I noticed it just now [19:42] good advice matt, that looks much nicer [19:57] sweet [19:57] Hm, that download location menu shoots off the edge of the white area quite funkily in Gecko [19:57] yes it does [19:58] And it's all Lund University's fault! [19:58] Well, not entirely [19:58] I do truncate it [19:58] United States Northeastern University have a lot to answer for, too [19:59] I'm thinking about switching to just choosing a country [20:00] yes, random + weighted in favor of reported bandwidth [20:00] so someone with twice the bandwidth will get aprox twice the traffic [20:01] mpt: you missed it, I said: so someone with twice the bandwidth will get aprox twice the traffic [20:05] ah cool [20:06] I guess the only downside would be for countries with large dimensions, i.e. Russia, USA, and to some extent Australia [20:06] USA has good bandwidth throughout the country [20:06] true [20:06] I don't know about Russia and Australia though [20:07] Africa is another tricky one as I understand it... there are few mirrors and rules for picking the best are complex [20:07] That sounds like a job for a learning algorithm :-) [20:07] ah, indeed [20:08] but doing the measurements would require doing dummy downloads every so often [20:08] well, you could learn from user's choices [20:09] and gauge success if they don't return for a second attempt [20:10] Oh, I wasn't thinking you would still let them choose [20:10] oh, I see [20:11] rather choose automatically based on how fast each mirror had been recently for people from that IP block or whatever [20:12] hmm, now I think about it, that would be really quite hard [20:12] yeah [20:12] unless you have geographically distributed download-testing machines, which is probably overkill :-) [20:12] I dn't have that [20:13] ssh timbuktu.canonical.com === apachelogger_ is now known as apachelogger