Perl 6 - the future is here, just unevenly distributed

IRC log for #rosettacode, 2012-01-06

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

All times shown according to UTC.

Time Nick Message
00:22 mwn3d_phone1 joined #rosettacode
00:33 kpreid joined #rosettacode
00:42 kpreid joined #rosettacode
01:24 nimdAHK joined #rosettacode
03:09 deltree_ joined #rosettacode
03:15 r4m joined #rosettacode
03:52 deltree__ joined #rosettacode
04:45 mwn3d_phone joined #rosettacode
05:03 Coderjoe mikemol: btw, are squid and mw set up properly to allow mw to tell squid to purge pages when needed?
05:15 mikemol Coderjoe: Yes.
05:21 mikemol Mm. Looks like squid doesn't currently have the capability of compressing objects in transit.
05:33 mwn3d_phone mikemol: http://wiki.squid-cache.org/Features/Gzip?highlight=%28%5CbCategoryFeature%5Cb%29%7C%28faqlisted.yes%29
05:33 fedaykin "Features/Gzip - Squid Web Proxy Wiki" http://rldn.net/wH3q
05:33 mwn3d_phone That's as good as it gets I guess
05:38 eMBee or use adifferent proxy line nginx
05:48 maustin joined #rosettacode
05:49 eMBee http://tumblr.intranation.com/post/766288369/using-nginx-reverse-proxy
05:50 eMBee https://calomel.org/nginx.html (scroll down)
05:50 eMBee http://wiki.nginx.org/HttpGzipModule
05:50 eMBee seems doable
05:51 eMBee http://wiki.nginx.org/ReverseProxyCachingExample
06:34 mikemol If it accept's MediaWiki's purge comands, perhaps. But even then only as part of some major future upgrade, which (based on pattern) will probably be a year or two down the road.
06:36 mikemol I did chat with the prgmr folks, and they're looking at upping the bandwidth caps. So serverside is not likely to be an issue, long-term. If the bandwidth consumed is sufficiently significant to impact mobile users, I might re-enable apache's content-encoding awareness and eat it in cache misses.
06:37 mikemol The accelerator proxy role is one of the parts of the site that it's really not possible to poke and tweak without downtime. Even a "/etc/init.d/squid reload" tends to break things until I execute a "/etc/init.d/squid restart".
06:37 mikemol So I really don't want to poke it unless something forces the issue.
06:40 mikemol Now, an intermediary stage might be setting up something like nginx on port 80, have the squid accelerator as a parent cache, and have nginx apply the compression, but not have a cache (or a cache of only a few seconds). That could be at least dry-tested by putting ngingx on a different port.
08:06 mischi joined #rosettacode
08:36 r4m joined #rosettacode
09:47 eMBee mikemol: but testing an alternative proxy can be done without downtime if you let the proxy run on a different port. so you keep running squid until the new one works and then you switch ports.
09:47 eMBee oops, that's just what you said last, started writing to fast ... :-)
09:48 eMBee but then i was thinking of having nginx talk directly to mediawiki like squid
09:49 eMBee i would not nest them
09:50 eMBee just put nginx on port 8000 or whatever and ask the active users to test if it is working. i have rc open allthe time, i'd just need to enter the other port once
10:00 achromic joined #rosettacode
11:23 mischi joined #rosettacode
15:31 mikemol eMBee: the complication is in whether nginx would support MediaWiki's means of telling it to purge old pages.
15:32 mikemol MW is written upstream under the assumption that the proxy server, if any, will be squid. I don't know how much that constrains me.
15:33 eMBee ah, but that can be tested even if nginx is running on another port
16:43 mwn3d_phone1 joined #rosettacode
17:26 r4m joined #rosettacode
17:35 mwn3d_phone1 joined #rosettacode
19:44 mischi joined #rosettacode
22:24 r4m joined #rosettacode
23:45 mwn3d_phone1 joined #rosettacode

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