Perl 6 - the future is here, just unevenly distributed

IRC log for #rosettacode, 2013-03-27

| Channels | #rosettacode index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:35 mwn3d_phone1 joined #rosettacode
02:57 kpreid joined #rosettacode
03:02 kpreid joined #rosettacode
04:55 BenBE_ joined #rosettacode
07:42 mischi joined #rosettacode
09:20 sivoais joined #rosettacode
09:29 sivoais joined #rosettacode
09:39 sivoais joined #rosettacode
09:41 mischi joined #rosettacode
09:48 sivoais joined #rosettacode
09:58 sivoais joined #rosettacode
10:08 sivoais joined #rosettacode
10:17 sivoais joined #rosettacode
10:27 sivoais joined #rosettacode
10:36 sivoais joined #rosettacode
10:46 sivoais joined #rosettacode
10:55 sivoais joined #rosettacode
11:05 sivoais joined #rosettacode
11:15 sivoais joined #rosettacode
11:20 sirdancealo2 joined #rosettacode
11:24 sivoais joined #rosettacode
11:34 sivoais joined #rosettacode
11:39 sirdancealo2 joined #rosettacode
11:43 sivoais joined #rosettacode
11:53 sivoais joined #rosettacode
11:54 mwn3d_phone1 joined #rosettacode
12:02 sivoais joined #rosettacode
12:05 sirdancealo2 joined #rosettacode
12:10 mwn3d_phone joined #rosettacode
12:12 sivoais joined #rosettacode
12:21 sivoais joined #rosettacode
12:26 sirdancealo2 joined #rosettacode
12:30 sivoais joined #rosettacode
12:40 sivoais joined #rosettacode
12:47 sirdancealo2 joined #rosettacode
12:50 sivoais joined #rosettacode
12:59 sivoais joined #rosettacode
13:09 sivoais joined #rosettacode
13:10 sirdancealot joined #rosettacode
13:19 sivoais joined #rosettacode
13:29 sivoais joined #rosettacode
13:36 sivoais joined #rosettacode
14:57 mikemol OK, need volunteers.
14:58 mikemol Every page that takes more than 1s to render (on average) needs to be split into subpages. Wholesale splitting by alphabetical/lexical grouping is probably appropriate.
14:58 mikemol The page render times are rediculous, and the the pages change too frequently to effectively cache.
14:59 mikemol Feel free to point out these statements in the IRC logs, if necessary. I've got to get back to work.
15:11 mikemol Either mwn3d_phone or myself can have Google Analytics spit out a list of pages with high average load times. But I can't do it myself until tomorrow evening, at the earliest.
15:37 mwn3d_phone Top five according to avg page load time over the last three weeks: line printer, terminal control/clear, longest common subsequence, SEDOLs, send email
15:37 mwn3d_phone Send email is the fastest with 44s
15:38 mwn3d_phone Pretty snappy for me...
15:38 mwn3d_phone I must be getting the info out of analytics wrong
15:38 mwn3d_phone I'll try again after lunch
15:46 mikemol mwn3d_phone: the analytics are going to average out a thousand or so samples over the period being looked at. Caching is going to make single-source analysis unreliable and/or expensive.
15:52 mikemol Also, try focusing on the pages following the march 11 switchover to Cloudflare.
15:54 mikemol The signal will be noisier, but I expect CF pulls a page once per datacenter, rather than just once. A page they've got cached in Hong Kong won't be cached in London.
15:55 mikemol The hit/miss ratio will thus vary based on geographic demand, rather than just whether or not anyone had requested the object previously.
15:55 * mikemol may no longer work at a CDN, but it at least let him get some perspective on geo-scale architectures...
15:57 cecil` joined #rosettacode
16:45 sirdancealo2 joined #rosettacode
17:13 mwn3d_phone Now that I look closer google analytics is probably a pretty terrible source for this. The line printer page has an average load time of 81s since march 11 with a sample size of one page load.
17:13 mikemol GA is the only source for it. We don't get that kind of data from CF.
17:13 mwn3d_phone Oi
17:14 mikemol 81s for that page sounds plausible, depending on what day it was, and whether there was a load spike.
17:15 mikemol It's an infrequently-accessed page, so it's not going to be cached much. But it's got a lot of content, and it changes from time to time.
17:15 mwn3d_phone Ok well I added an advanced filter to only give pages with more than 15 page loads
17:15 mwn3d_phone Now the top five are: n-queens, quicksort, bubble sort, X11 window creation, walk a directory recursively
17:16 mikemol You'd also be amazed at how far GA differs from the data I *do* see from CF. Spambot attempts and random crawlers appear in the CF data, but not in GA.
17:17 mikemol We're talking differences in visit count in excess of 100%...
17:18 mwn3d_phone I guess its usually good to have the spambots filtered out
17:18 mikemol It's catching a lot of that. At least a few hundred spam visits per day.
17:18 mikemol Obviously, a lot is still getting through.
17:18 mwn3d_phone Heh...looked further down the list on GA...the main page has a 5s load time
17:19 mikemol Yeah, those templates that pull from Semantic MediaWiki make expensive queries.
17:20 mwn3d_phone The last thing to load for me is usually the g+/tiwtter/facebook buttons
17:20 * mikemol nods
17:20 mikemol Those are now async, though. CloudFlare is doing some fancy page rewriting for us in that area.
17:21 mwn3d_phone I wonder if we could just go off of [[Special:LongPages]]
17:21 fedaykin http://rosettacode.org/wiki/Special:LongPages
17:21 mikemol Bruce Schneir's blog just implemented something I thought was kinda cool. They've got this sharethis-style widget that requires you to click on it before it loads content.
17:22 mikemol The idea is privacy, but I have to think that would dramatically improve page load speed.
17:22 Hypftier Didn't I recommend that months ago already? (on the grounds of privacy, though) ;)
17:23 Hypftier But is page load speed really that tied to a few images from external sites?
17:26 mwn3d_phone It might just be another way to chip away at it
17:30 mikemol Hypftier: Don't remember.
17:31 mikemol Hypftier: And, honestly, I don't care terribly much about privacy in the context of RC. Those who demand it will implement effective countermeasures on their side. Askins servers to do it is like running a "default allow, deny specific" firewall; leaky as a sieve.
17:31 mwn3d_phone joined #rosettacode
17:51 mikemol Hypftier: To be clear, my original statements on privacy still hold; RC will not share user email addresses. I won't speak to source IP addresses yet, as that could easily paint me into a corner wrt combatting spam.
17:52 mikemol I.e., at some point, I'd like to use the email addresses of spam accounts to identify the service they're using to create temporary email accounts, and then use the source IP addresses to help those email service providers block the malicious users from using their services.
17:53 mikemol And for whoever the jerk is monitoring this channel and adapting his spam measures, go ahead and implement a workaround for that. It amuses me to cause you work in that way simply by musing about plausible means of implementing spam countermeasures.
18:18 cecil` I just started with Scala and found rosettacode. What are things that need a Scala example and are not very hard to do? (Start small.)
18:18 mikemol cecil`: Hang on
18:19 mikemol http://rosettacode.org/wiki/Repor​ts:Tasks_not_implemented_in_Scala
18:19 fedaykin "Reports:Tasks not implemented in Scala - Rosetta Code" http://rldn.net/16edC
18:19 mikemol That's the full list.
18:19 mikemol I'm not qualified to be able to definitively tell you what's easy or difficult to do in Scala.
18:21 cecil` Well the first two I did from that list (100 doors and ackerman) where already implemented. (But not very efficient.)
21:58 sirdancealo2 joined #rosettacode
22:18 mwn3d_phone1 joined #rosettacode
22:43 mwn3d_phone joined #rosettacode
22:48 mwn3d_phone1 joined #rosettacode
23:00 mwn3d_phone joined #rosettacode
23:02 mwn3d_phone1 joined #rosettacode

| Channels | #rosettacode index | Today | | Search | Google Search | Plain-Text | summary