| 00:04 | <Hixie> | TabAtkins: thanks, awesome. |
| 00:06 | <zewt> | Ian Hickson to whatwg, whatwg <- heh |
| 00:24 | <Hixie> | ok the wiki login problems from earlier should be fixed |
| 00:24 | <Hixie> | (/tmp was full; we ended up moving it to a different partition with orders of magnitude more space) |
| 00:48 | <AryehGregor> | Can I run an AM3 CPU in my AM2 motherboard? |
| 00:48 | <AryehGregor> | Wikipedia claims yes . . . |
| 00:54 | <aho> | the mobo vendor page should help (you gotta go there for the new bios anyways) |
| 00:54 | AryehGregor | finds that for about $100, he can upgrade his Athlon 64 X2 5000+, which is 2.6 GHz dual-core, to an Athlon II X4 645, which is 3.1 GHz quad-core |
| 00:55 | <AryehGregor> | But I'm possibly only using this computer for another few months. Hmm. |
| 00:55 | <aho> | i still got a x2 4200+ here |
| 00:56 | <AryehGregor> | Thing is, my tests are taking 30 minutes to run and increasing. |
| 00:56 | <aho> | heh |
| 00:57 | <aho> | on my old machine, the first png re-compression run of inkscape.org's pngs took over 2 days :> |
| 00:57 | <aho> | that was fun, i tell you :) |
| 00:59 | <aho> | http://inkscape.org/crunchman/inkscape_web.cmlog.html |
| 00:59 | <aho> | totally worth it though :> |
| 00:59 | <zewt> | are your tests running in parallel? heh |
| 01:01 | <AryehGregor> | I have two cores and I'm running them in two browsers at once, so yeah, I suppose so. |
| 01:01 | AryehGregor | finds he has pretty much no upgrade options for his laptop's CPU. Feh. |
| 01:03 | <AryehGregor> | Seems like CPUs that are targeted at laptops and draw 35W are a lot slower than ones targeted at desktops that draw 95W, and more expensive. Who'd've thought? |
| 01:03 | <AryehGregor> | Oh well. |
| 01:13 | timeless | chuckles |
| 01:14 | <timeless> | AryehGregor: have you considered rent-a-cloud? |
| 01:20 | <zewt> | heh |
| 01:20 | <zewt> | i thought about suggesting EC2, but it'd probably need to be a bit more than that to be worth the bother |
| 01:22 | <zewt> | between the bother of setting it up, and the time overhead of spinning up instances |
| 06:30 | <hsivonen> | jgraham: does Ragnarök clone Firefox's tree depth limit that's not in the parsing spec? |
| 06:30 | <hsivonen> | jgraham: I'm rather surprised that http://www.documentacatholicaomnia.eu/1815-1875,_Migne,_Patrologia_Latina_01._Rerum_Conspectus_Pro_Tomis_Ordinatus,_MLT.html fails in Ragnarök the way it fails in Firefox (though with different text color) |
| 07:11 | <jgraham> | hsivonen: I thought we had just implemented the spec there |
| 07:13 | <jgraham> | Oh, wait, if you mean total tree depth |
| 07:13 | <jgraham> | Yeah, we have a limit on how deep trees can be |
| 07:14 | <jgraham> | iirc our final soluion was quite close to the firefox behaviour |
| 07:14 | <hsivonen> | jgraham: thanks |
| 08:19 | <annevk> | abarth, seems like the IRI WG is still not capable of delivering a better URL specification :( |
| 08:20 | <abarth> | annevk: big surprise :) |
| 08:22 | <annevk> | Yeah, I would have been surprised if they did, actually |
| 08:29 | <erlehmann> | better in what regard? |
| 08:30 | <annevk> | One that defines what you have to implement |
| 08:37 | <hsivonen> | another day, another threat against https :-( |
| 08:42 | <annevk> | Pointer hsivonen? |
| 08:50 | <annevk> | http://www.usenix.org/events/sec11/stream/jackson/index.html interesting history on postMessage() |
| 08:51 | <annevk> | Especially on how browser vendors were played by security researchers (prolly just abarth and collin :) ) |
| 08:51 | <erlehmann> | annevk, this? http://www.theregister.co.uk/2011/09/19/beast_exploits_paypal_ssl/ |
| 08:51 | <abarth> | lies! |
| 08:54 | <hsivonen> | annevk: what erlehmann said |
| 08:59 | <annevk> | Hmm, Yngve now and then tells me about silly websites still using less-than-the-latest encryption stuff |
| 08:59 | <annevk> | guess he had a point :( |
| 09:07 | <MikeSmith> | "Although versions 1.1 and 1.2 of TLS aren't susceptible, they remain almost entirely unsupported in browsers and websites alike" |
| 09:07 | <MikeSmith> | huh? |
| 09:08 | <MikeSmith> | oh |
| 09:11 | hsivonen | is amused by getting garbled email from an IETF old-timer |
| 09:25 | <MikeSmith> | http://news.cnet.com/8301-30685_3-20108633-264/researchers-to-detail-hole-in-web-encryption/ |
| 09:25 | <MikeSmith> | However, Adam Langley, a TLS advocate and expert at Google, isn't concerned. |
| 09:25 | <MikeSmith> | "The researchers disclosed BEAST to browsers so I'm not going to comment in detail until public. It's neat, but not something to worry about," Langley tweeted yesterday. |
| 09:51 | annevk | wonders why there is so much fail going on in http://w3c-test.org/webapps/ProgressEvents/tests/submissions/Ms2ger/interface.html |
| 10:00 | <annevk> | Ms2ger, no idea if you are reading this, but I am making some more ProgressEvent tests |
| 10:11 | <annevk> | Browsers fail http://w3c-test.org/webapps/ProgressEvents/tests/submissions/Ms2ger/ |
| 10:12 | <annevk> | jgraham, even for synchronous callbacks you need async tests right? |
| 10:14 | <jgraham> | annevk: I don't really understand the question, but if in doubt async tests always work |
| 10:49 | <Philip`> | zewt: Usually EC2 only seems to take about five minutes to start an instance, and setting it up seems pretty straightforward when you follow tutorials, so I wouldn't have thought it's too much bother for a process that would otherwise take 30 minutes |
| 11:01 | <karlcow> | http://www.w3.org/wiki/Html-data-tf |
| 11:03 | <annevk> | jgraham, addEventListener("test", function() { something }) within a synchronous test is that okay? |
| 11:03 | <annevk> | the event fires synchronously |
| 11:11 | <jgraham> | annevk: No, because exceptions from the event listener won't propogate to the enclosing function (I assume) |
| 11:23 | <zcorpan> | annevk: use t.step_func |
| 12:33 | <asmodai> | boom, and DigiNotar just went tits up. |
| 12:38 | <asmodai> | Mmm, and a TLS1.0 hacking/decrypting implementation: http://www.theregister.co.uk/2011/09/19/beast_exploits_paypal_ssl/ |
| 13:29 | <hsivonen> | asmodai: URL to latest DigiNotar news? |
| 13:31 | <asmodai> | hsivonen: let me see if I can find an English one |
| 13:31 | <asmodai> | hsivonen: unless you can read Dutch? |
| 13:32 | <asmodai> | http://www.itpro.co.uk/636244/diginotar-goes-bankrupt-after-hack |
| 13:34 | <hsivonen> | asmodai: thanks. (I can't read Dutch without Google Translate) |
| 13:35 | <asmodai> | :) |
| 13:41 | <hsivonen> | is there some kind of requirement to always say that HTML5 is not going to kill Flash? http://html5doctor.com/html5-briefing-notes-journalists-analysts/ |
| 13:45 | <asmodai> | I think Flash will lose some market share in that department, unavoidable |
| 14:29 | <annevk> | zcorpan, but you need an async test right? |
| 14:35 | <zcorpan> | annevk: yeah |
| 15:01 | <espadrine> | Hello, I'm trying to summarize what wonderful new stuff comes with upcoming updates to specs, over at https://github.com/espadrine/New-In-A-Spec . I have currently covered DOM4 and ES6. I want to make it easy for anyone to see a glimpse of the future. Tell me what you think, offer pull requests, etc. |
| 15:06 | <bga_> | oh @x == p@x >_< |
| 16:31 | <Ms2ger> | annevk, yeah, WebIDL stuff is full of fail |
| 16:32 | <annevk> | in browsers? |
| 16:32 | <Ms2ger> | Yeah |
| 16:44 | <AryehGregor> | timeless, rent-a-cloud isn't much good for running tests that only work in browsers, unless I want to set up a headless desktop and log in remotely . . . doesn't seem worth the effort. |
| 16:45 | <timeless> | what's wrong w/ that approach? |
| 16:46 | <timeless> | if you configure things magically, you can skip the login bit too |
| 16:46 | <timeless> | an agent which gets orders from a webpage |
| 16:46 | <AryehGregor> | That's a clever idea. |
| 16:46 | <AryehGregor> | Maybe I'll pursue it at some later date. |
| 16:47 | <AryehGregor> | But for now I think I'll just recognize that I'm going to keep using a desktop for the foreseeable future, not just a laptop, and buy a decent CPU. |
| 16:47 | <timeless> | also a good plan :) |
| 16:51 | <Ms2ger> | Oh look, it's ChrisWilson |
| 16:51 | <ChrisWilson> | heh. auto-reconnect |
| 16:51 | <timeless> | heh |
| 16:52 | ChrisWilson | is at W3C Web-and-TV meeting, trapped in room all day |
| 16:52 | timeless | reaches the point in http://www.usenix.org/events/sec11/stream/jackson/index.html where he thanks IE for breaking Mixed Content |
| 16:52 | timeless | wants to thank the IE team for doing that |
| 16:53 | <timeless> | well, he didn't really thank them so much as praise them, but.. |
| 16:55 | <AryehGregor> | So it looks like AM3 CPUs are not officially supported by ASUS on the M2N. |
| 16:58 | <AryehGregor> | More than that, the latest M2N BIOS was released April 2009, and AM3 was launched in February 2009, so I'm pretty sure there's no support. |
| 16:58 | <AryehGregor> | Grr. |
| 17:07 | <Philip`> | AryehGregor: At least with Windows on EC2, you just launch it and wait a few minutes then rdesktop into it and run whatever you fancy |
| 17:07 | <AryehGregor> | Interesting thought. |
| 17:08 | <AryehGregor> | I could even run IE that way instead of having to steal my parents' laptop. |
| 17:08 | <AryehGregor> | Hmm. |
| 17:16 | <AryehGregor> | "One EC2 Compute Unit provides the equivalent CPU capacity of a 1.0-1.2 GHz 2007 Opteron or 2007 Xeon processor." |
| 17:16 | <AryehGregor> | That doesn't sound like a lot. |
| 17:17 | <AryehGregor> | But the high-memory extra large instance is two virtual cores with 3.25 compute units each, and 17.1 GB of RAM. |
| 17:17 | <AryehGregor> | $0.62 per hour for Windows. |
| 17:17 | <AryehGregor> | Double extra large is still only $1.24 per hour. |
| 17:17 | <AryehGregor> | For using them only a few hours a day, that sounds like an awfully good deal. |
| 17:18 | <Philip`> | High-CPU sounds better than high-memory, I would have thought |
| 17:19 | <Philip`> | (They round up to an integer number of hours, by the way) |
| 17:19 | <AryehGregor> | The high-CPU medium instance only has 1.7G of memory, which isn't very good for running large test pages in multiple browsers. The high-CPU extra large has 7G of memory, but eight cores with 2.5 compute units each, which is worse for my needs than two or four cores at 3.25 compute units each. |
| 17:19 | <AryehGregor> | Good to know. |
| 17:19 | AryehGregor | calculates that this is likely to cost <$500 a year |
| 17:20 | <AryehGregor> | When I calculated it for my server, it didn't look so good, but of course servers have to be running all the time. |
| 17:21 | <AryehGregor> | At least mine did. |
| 17:21 | <AryehGregor> | Okay, I should probably set one of these up. |
| 17:27 | <Philip`> | (http://docs.amazonwebservices.com/AWSEC2/2011-07-15/GettingStartedGuide/ seems like the relevant documentation, if you haven't seen that already) |
| 17:28 | <Philip`> | (and presumbly you'd want to tell it to launch an instance based on http://aws.amazon.com/amis/Microsoft-Windows/7454225207018905 or similar) |
| 17:29 | <AryehGregor> | Thanks! |
| 17:29 | AryehGregor | looks |
| 17:38 | <AryehGregor> | Philip`, if I want to not have to reinstall the browsers every time, I'm going to need some block storage too, right? |
| 17:38 | <AryehGregor> | How does that work? |
| 17:42 | <Philip`> | AryehGregor: If you create an EBS-based instance (not the old-fashioned S3-based style), then it creates you a new EBS root volume by cloning the original image, and changes to that volume will persist until you terminate the instance (or it suffers a failure) |
| 17:42 | <AryehGregor> | How do I do that? Does it have any other practical implications? |
| 17:42 | <Philip`> | "terminate" is different from "stop" - you can do the latter and it'll keep the volumes and won't charge you for running the machine |
| 17:43 | <Philip`> | (except for the cost of storage, which is fairly trivial) |
| 17:43 | <AryehGregor> | Oh! |
| 17:43 | <AryehGregor> | I can stop my instance and it will just hibernate it or something, and only charge me for storage? |
| 17:44 | <zewt> | well, if you want to be able to spawn a bunch of temporary instances, then you probably want to set up a snapshot, and start instances with only instance storage and no EBS storage |
| 17:44 | <zewt> | if you want to be able to parallelize tests over a bunch of instances |
| 17:44 | <AryehGregor> | I can parallelize over cores just as easily. |
| 17:44 | <AryehGregor> | Or more easily, probably. |
| 17:45 | <Philip`> | AryehGregor: You just make sure the AMI (image) states "backed by Amazon EBS" or similar, which all Windows 2008+ do (according to http://docs.amazonwebservices.com/AWSEC2/latest/UserGuide/index.html?Concepts_BootFromEBS.html) |
| 17:45 | <zewt> | you'd be limited by the number of cores you can have on a single instance, though (which is 8, I think) |
| 17:45 | <AryehGregor> | Okay, thanks. |
| 17:46 | <AryehGregor> | zewt, I'm not going to want to run more than about two or three copies of gentest.html at once, and it only runs in Gecko/WebKit. |
| 17:47 | <zewt> | i'd think if you have so many tests that it takes half an hour to run, you'd want a framework that can parallelize arbitrarily |
| 17:47 | <Philip`> | AryehGregor: Yeah, if you stop it then it only costs storage, though I think there's no strong guarantees about data persistence so you'd probably want to be more careful (backup to S3 etc) if it takes hours to set up |
| 17:47 | <zewt> | (not that that implies the framework actually can, of course :) |
| 17:48 | <zewt> | backing up an instance to S3 is pretty trivial, i have a server doing it automatically |
| 17:49 | AryehGregor | cannot believe how much of a pain it is to use IE8 for even something as simple as just installing Chrome |
| 17:49 | <zewt> | wget.exe? :) |
| 17:49 | <ChrisWilson> | so you're saying IE is no longer the browser of choice for installing another browser? |
| 17:49 | <AryehGregor> | It doesn't help that this copy of Windows is Windows Server and I'm trying to use it as a desktop. |
| 17:50 | <AryehGregor> | It seems to have JS disabled by default for most websites, including bing.com. |
| 17:50 | <AryehGregor> | Meaning of course that it prompts me to override it on every website, which I do. |
| 17:50 | <Philip`> | Windows Server's IE is certainly a little paranoid |
| 17:50 | <zewt> | advance EC2 tip: when it says "shutting down an instance doesn't delete the EBS storage associated with it", they're lying. |
| 17:50 | <AryehGregor> | I tried to figure out how to turn it off but couldn't. |
| 17:51 | <zewt> | (creating an instance with EBS storage is different than creating an instance and EBS storage separately and attaching it--the former *will* silently delete your data, despite what they say) |
| 17:51 | <Philip`> | (In cases other than when you Terminate and haven't disabled DeleteOnTermination on the volume?) |
| 17:51 | <AryehGregor> | It doesn't seem to let me download anything . . . |
| 17:52 | <Philip`> | AryehGregor: Downloading could be very dangerous, why should it let you do that? |
| 17:52 | <zewt> | ("delete on termination" isn't something visible from the UI, at least not in any way I saw when I was setting this stuff up a while back) |
| 17:52 | <Philip`> | If I remember correctly it should give you an option to whitelist sites for downloads |
| 17:52 | <Philip`> | zewt: Yeah, I think I needed command-line tools to toggle it, which is silly |
| 17:53 | <zewt> | of course, I played around with it for a couple days to see what the gotchas are, which I was quickly thankful for |
| 17:53 | <zewt> | Philip`: the problem is with the UI saying "we don't delete your data! please wait, deleting your data..." D: |
| 17:54 | AryehGregor | Googles for how to turn it off, succeeds |
| 17:55 | <Philip`> | Non-DeleteOnTermination isn't particularly useful since I think you can't actually boot from the non-deleted volume, you can only attach it to an existing instance |
| 17:55 | <Philip`> | (though I may be mistaken) |
| 17:55 | <AryehGregor> | Ironically, one of the first things I'm using Chrome for is downloading the latest IE, because it's too painful to do in IE8. |
| 17:55 | <zewt> | well, i'm pretty sure you can attach a volume and boot from it, though I don't recall the specifics off-hand |
| 17:56 | <zewt> | i recall having an instance that wouldn't boot, and i was able to dismount the boot drive, mount it in a temporary instance, debug the problem and put it back in its own instance |
| 17:56 | <AryehGregor> | How stable is Firefox aurora vs. nightly? |
| 17:56 | <zewt> | which is very cool, since that's the sort of thing you normally have to file a ticket, wait hours and hope whoever's at the desk at the DC that day doesn't clobber your stuff |
| 17:58 | <smaug____> | AryehGregor: well, API-vise Aurora is pretty stable. Nightly may get some changes which are removed if there are regressions |
| 17:58 | <smaug____> | AryehGregor: but otherwise Nightly should be pretty stable |
| 17:58 | <AryehGregor> | smaug____, I meant in terms of "if I'm only using it for running spec-related tests, should I use nightly or Aurora?" |
| 17:58 | <smaug____> | use Nightly |
| 17:59 | <smaug____> | especially for Web DOM Core, DOM4, DOM-foobar-whateveritiscalled |
| 17:59 | <Ms2ger> | DOM4 today |
| 17:59 | <smaug____> | We need to change it. It has been DOM4 way too long |
| 17:59 | <zewt> | heh, i was really surprised there wasn't huge pushback from the DOM3 people for that |
| 18:00 | <Ms2ger> | DOM3 people? Are any still around? |
| 18:00 | <AryehGregor> | "Internet Explorer 9 is now installed |
| 18:00 | <AryehGregor> | Some Internet Explorer files were in use during setup. Restart your computer to use Internet Explorer 9." |
| 18:00 | <AryehGregor> | . . . |
| 18:00 | <Ms2ger> | Yeah |
| 18:00 | <AryehGregor> | Hmm, wait. |
| 18:01 | <AryehGregor> | Does restarting work? |
| 18:01 | <AryehGregor> | I mean, if I tell the OS to restart, Amazon doesn't know to kill my storage, right? |
| 18:01 | <zewt> | well, doug schepers |
| 18:01 | <smaug____> | PLH is very much still around |
| 18:01 | <AryehGregor> | smaug____, okay, I'll use nightly. |
| 18:01 | <Ms2ger> | Granted |
| 18:01 | <Ms2ger> | And besides them? |
| 18:01 | <AryehGregor> | I should probably use Chrome Canary too. |
| 18:02 | <AryehGregor> | Ms2ger, is this going to be like the Life of Brian "What have the Romans done for us?" scene? |
| 18:02 | <Ms2ger> | Dammit |
| 18:02 | <shepazu> | what about "DOM3 people"? |
| 18:02 | <smaug____> | zewt: shepazu wasn't around when DOM3 Core was written, IIRC |
| 18:02 | <Ms2ger> | OTOH, shepazu already worked on DOM4 |
| 18:02 | <zewt> | he's an editor on DOM3 Events |
| 18:03 | <shepazu> | PLH was, and he's a manager at W3C now |
| 18:03 | <shepazu> | yeah, DOM3 Events has taken way too long |
| 18:03 | <zewt> | and it seemed he wasn't happy with DOM-whatever-you-call-it-Core's events spec sort of stepping on its spec turf |
| 18:03 | <shepazu> | mostly because it's been hard to get anyone from the browsers to help out |
| 18:03 | <zewt> | (just my impression) |
| 18:03 | <Philip`> | AryehGregor: Reboot should work; in the worst case (normal shutdown) it's like "stop", not "terminate", so it shouldn't delete storage |
| 18:04 | <AryehGregor> | Philip`, and if I right-click the instance and click "Stop", it will stop billing me for everything but storage and be restartable later? |
| 18:04 | <AryehGregor> | Is this stuff documented somewhere? |
| 18:04 | AryehGregor | was looking but didn't see any sufficiently obvious links |
| 18:04 | <Ms2ger> | Oh, and darobin |
| 18:04 | <zewt> | you'll always be billed for storage |
| 18:05 | <Ms2ger> | And Björn |
| 18:05 | <shepazu> | zewt: no, I objected to the deliberate conflicts between specs without an effort to address the issues in DOM3 Events first…. please stop spreading rumors and misinformation |
| 18:05 | <zewt> | i'm not spreading anything, and i'm explicitly labelling my impressions as such |
| 18:05 | <Philip`> | http://support.rightscale.com/06-FAQs/FAQ_0149_-_What's_the_difference_between_Terminating_and_Stopping_an_EC2_Instance%3F#Stop_Instance seems informative |
| 18:06 | <shepazu> | yeah, zewt was against there being any DOM specs |
| 18:06 | <shepazu> | (just my impression) |
| 18:06 | <smaug____> | :) |
| 18:06 | <Ms2ger> | shepazu, so what are your impressions about me? :) |
| 18:06 | <zewt> | shepazu: you can draw what impressions you like :) |
| 18:08 | <shepazu> | if you can't see that kind of broad misstatement as rumor-mongering, I can't convince you |
| 18:08 | <Philip`> | AryehGregor: Also http://docs.amazonwebservices.com/AWSEC2/latest/UserGuide/index.html?Concepts_BootFromEBS.html#Stop_Start |
| 18:08 | <zewt> | to be fair, if I was the one writing the events spec in dom-core, I'd have separated the "writing the spec" and "lobbying for changes" part; speccing something that was both a separate spec *and* effectively lobbied for changes to what was specced probably made things harder than they could have been |
| 18:09 | <zewt> | but that's in retrospect, of course |
| 18:10 | <shepazu> | no, it just wasn't done respectfully, in my opinion… it could have been, but anne chose not to |
| 18:11 | shepazu | off to another teleconference... |
| 18:12 | <zewt> | shepazu: and stating my honest impressions based on following the discussions (fairly closely) is not a "misstatement" or "rumor-mongering" |
| 18:17 | <zewt> | AryehGregor: might want to run a quick benchmark on your instance vs. your own machine to see if it's actually faster per core, before spending too long setting a lot of stuff up |
| 18:18 | <zewt> | unless merely offloading the CPU is worth it for you, I guess |
| 18:24 | timeless | is glad to see Philip` and zewt seem to have convinced AryehGregor about the cloud approach to testing |
| 18:24 | <zewt> | well, it's not really the cloud approach with only one server :) |
| 18:25 | Philip` | finds it more useful as a way to avoid having to have a Windows machine of his own :-) |
| 18:25 | <zewt> | ew, i'd never run windows on a remote system |
| 18:26 | <zewt> | vmware would be much better for that |
| 18:26 | <Philip`> | (for compiling and stuff) |
| 18:27 | <zewt> | well, at least for compiling, you can in theory do msvc builds from a terminal--though i've never messed with that |
| 18:28 | <Philip`> | Yeah, I use vcbuild to do it automatedly, not the IDE |
| 18:28 | <Philip`> | though it doesn't work quite the same |
| 19:09 | <GlitchMr> | Personally, I would like to see way of making form fields disable after choosing certain radio buttons or checkboxes... but yeah... Web Forms is already big part of HTML5 specification... |
| 19:30 | <AryehGregor> | Do Firefox nightlies auto-update on Windows? |
| 19:31 | <GlitchMr> | AryehGregor, probably not :P. |
| 19:31 | <Ms2ger> | Yes |
| 19:31 | <timeless_> | AryehGregor: yes |
| 19:31 | <timeless_> | daily |
| 19:31 | <GlitchMr> | http://kb.mozillazine.org/Software_update |
| 19:31 | <AryehGregor> | GlitchMr, I was mostly directly the question to the various Mozilla devs who hang out here. |
| 19:31 | <timeless> | you'll get an annoying dialog every morning |
| 19:31 | <AryehGregor> | Ms2ger, timeless, thanks. |
| 19:32 | <timeless> | and if it's your irc client, you'll have to disconnect daily |
| 19:32 | <timeless> | on mondays, you'll get 2 updates |
| 19:32 | <timeless> | on to update from friday to saturday (small) |
| 19:32 | <timeless> | s/on to/one to/ |
| 19:32 | <timeless> | and one to update from saturday to monday (full) |
| 19:32 | <timeless> | if you're really unlucky, you could get 3 updates |
| 19:33 | <timeless> | one to update friday to saturday, one to update saturday to sunday, and an hour later, one to update from sunday to monday :) |
| 19:33 | <timeless> | oh, and for extra restarts (and a chance for a bonus reboot), you can install Adobe Flash and Adobe Reader |
| 19:33 | <Ms2ger> | Unless you work during the weekend |
| 19:33 | <timeless> | (reader wants me to reboot) |
| 19:33 | <timeless> | oh, to do that, you need a "home pc", and this "internet to the home" thing |
| 19:33 | timeless | is kinda missing those |
| 19:34 | <AryehGregor> | timeless, this is why I don't use Firefox as my IRC client. |
| 19:35 | <AryehGregor> | I originally used Chatzilla but abandoned the idea very rapidly. |
| 19:35 | <timeless> | chatzilla can be hosted by xulrunner (discontinued) |
| 19:35 | <timeless> | but, those irc clients only work if your irc server |
| 19:35 | <timeless> | 's irc port is reachable from your computer |
| 19:35 | <timeless> | if not, you need something else (i use http) |
| 19:36 | <timeless> | hrm, IE has two modes in Metro... |
| 19:37 | <Philip`> | By "two", do you mean eight? |
| 19:37 | <timeless> | Windowed and Metro |
| 19:38 | <timeless> | i'd been using Windowed initially and didn't realize it was also a Metro app |
| 19:38 | <Philip`> | Are they orthogonal to the compatibility modes? |
| 19:38 | <timeless> | i tried loading http://google.com from explorer, and suddenly i got launched into Metro |
| 19:38 | <timeless> | i'd assume so |
| 19:38 | <timeless> | this is just a UI thing |
| 19:39 | <AryehGregor> | Time to generate tests in Chrome on my desktop: 27:48.740min. |
| 19:39 | <timeless> | http://www.favbrowser.com/internet-explorer-10-screenshots-and-video/ |
| 19:39 | <AryehGregor> | Time on my cloud VM: 2:56.945 min. |
| 19:39 | <AryehGregor> | Sounds like a win. |
| 19:40 | <Philip`> | Why so much difference? |
| 19:40 | <AryehGregor> | Presumably because my CPU is a piece of garbage. |
| 19:40 | <Philip`> | Is your desktop running on a Pentium 2 or something? |
| 19:41 | <AryehGregor> | No, an Athlon X2 64 5000+ or something. |
| 19:41 | <AryehGregor> | 2.6GHz. |
| 19:42 | <AryehGregor> | I don't know what the VM is using, but it's evidently better. |
| 19:43 | <timeless> | AryehGregor: can i have a thank=you? :) |
| 19:43 | <AryehGregor> | timeless, thank you! |
| 19:43 | <AryehGregor> | And to Philip` et al. |
| 19:43 | <Philip`> | Is it supposedly 3.25 x "a 1.0-1.2 GHz 2007 Opteron or 2007 Xeon processor"? |
| 19:43 | <timeless> | the cloud probably also has a much better hard disk, and probably network pipe :) |
| 19:44 | timeless | thanks Philip` and zewt for providing the actual guidance to get AryehGregor past the idea phase |
| 19:44 | <AryehGregor> | Philip`, yes. |
| 19:44 | <AryehGregor> | Interesting, the Firefox nightly still takes 12:55min. |
| 19:45 | <timeless> | :( |
| 19:45 | Philip` | wonders if the main problem with the X2 is that it has tiny L2 cache, because he doesn't know why else there'd be an order of magnitude difference |
| 19:45 | <timeless> | ActivePython-2.7.2.5-win64-x64.msi is not commonly downloaded and could harm yo |
| 19:46 | <timeless> | i supposed that IE10 is trying to be in with the 90s |
| 19:46 | <timeless> | `could harm yo` |
| 19:49 | <AryehGregor> | Are nightlies less optimized than Aurora, by any chance? |
| 19:51 | <roc> | shouldn't be |
| 19:51 | <roc> | what's the problem? |
| 19:51 | <AryehGregor> | Nothing, I was just surprised that Chrome seemed to show a ridiculously bigger speedup from my desktop to VM than Firefox did. |
| 19:51 | <AryehGregor> | I switched from Aurora on my desktop to Nightly on the VM, so I wondered if that might be related. |
| 19:52 | <AryehGregor> | (I didn't wind up using Chrome Canary because it sad-tabbed) |
| 19:53 | <roc> | no idea, but if you try Aurora on your VM and Nightly is a lot slower than it, I'd appreciate a bug report |
| 19:57 | AryehGregor | will fiddle |
| 19:57 | <timeless> | AryehGregor: fwiw, windows has some nice profiling stuff |
| 19:57 | <timeless> | if you're willing to do some extra work to see what's up |
| 19:57 | <AryehGregor> | I'm not really. |
| 19:57 | <timeless> | yeah, i didn't expect you were |
| 19:57 | <AryehGregor> | At least not right now. |
| 19:57 | <timeless> | if your vm doesn't have anything confidential in it, you might want to let someone clone it |
| 19:58 | timeless | isn't sure how clonable Amazon items are |
| 19:58 | <AryehGregor> | My VM has nothing useful in it whatsoever, it's a fresh Windows install with some browsers added. |
| 19:58 | <AryehGregor> | Anyone can trivially replicate it. |
| 20:03 | <AryehGregor> | There's no way to append text piece by piece to a textarea's value that's not O(n^2), is there? |
| 20:03 | <AryehGregor> | Or should I just append it to the DOM as a child of the textarea? |
| 20:03 | AryehGregor | tries that |
| 20:14 | <AryehGregor> | Hmm, now Chrome is back to taking a long time too. |
| 20:14 | <AryehGregor> | Maybe the earlier result was a fluke or bug or something. |
| 20:16 | <Philip`> | You should file a bug report: "Sometimes Chrome is too fast" |
| 20:21 | <GlitchMr> | ... |
| 20:21 | <GlitchMr> | On hardware made 7 years ago, Chrome is kinda slow... |
| 20:25 | <AryehGregor> | CPUs haven't gotten a lot faster in seven years. |
| 20:25 | <AryehGregor> | Certainly not ten times faster. |
| 20:26 | <aho> | hardware acceleration perhaps? |
| 20:27 | <Ms2ger> | Native HTML5 |
| 20:27 | <aho> | i was serious |
| 20:28 | <aho> | some drawing operations may be hardware accelerated on system A, but won't be hardware accelerated on system B |
| 20:28 | <AryehGregor> | 1) Hardware acceleration seems unlikely on a Windows Server VM. |
| 20:28 | <AryehGregor> | 2) Basically nothing is being painted here, if the browsers are being sane at all. |
| 20:29 | <aho> | sw rendering can be pretty fast if there is no ram->bus->vram copy stuff going on (also true in a headless environment) |
| 20:30 | <aho> | well, i have no clue what your tests are doing :) |
| 20:34 | <roc> | AryehGregor: maybe your tests ended early because of a bug? |
| 20:34 | <AryehGregor> | roc, yeah, that's what I suspect at this point. |
| 20:35 | <AryehGregor> | It seemed like a ridiculous speedup. |
| 20:35 | <AryehGregor> | The speedup I saw in Firefox was more like 2x, which is reasonable. |
| 20:35 | <AryehGregor> | I think I'll want to upgrade to the four-core version so I can run two sets of tests in parallel. |
| 20:36 | <roc> | about appending to a textarea ... of course the best way is to build up a JS string and append it to the textarea in one go, but presumably you can't do that for some reason |
| 20:37 | <AryehGregor> | I actually can. I don't think the constant factors are hurting me much here, though, given how much work I'm doing per test. |
| 20:37 | <AryehGregor> | Currently I'm using appendData() on the text node child, so it should be linear. |
| 20:38 | <roc> | I wouldn't bet on that |
| 20:38 | <AryehGregor> | Because the underlying storage might have to get reallocated regularly or something? |
| 20:38 | <AryehGregor> | Is there any easy way for me to get good profiling data on Firefox or Chrome? |
| 20:39 | <AryehGregor> | Hmm, Firebug works now, maybe that will help. |
| 20:39 | <AryehGregor> | Of course, it makes things take forever. |
| 20:41 | <AryehGregor> | I might have introduced an infinite loop somewhere here or something. |
| 20:46 | <AryehGregor> | Wow. |
| 20:47 | <AryehGregor> | Apparently building the string in pure JS made a ridiculous performance difference. |
| 20:47 | <AryehGregor> | Like over an order of magnitude compared to building it in the DOM. |
| 20:48 | <AryehGregor> | I have to double-check how much time my test framework is using here compared to the actual tests themselves. |
| 20:49 | <AryehGregor> | Firefox and Chrome now both take ~1min on the VM. |
| 20:50 | AryehGregor | has to check that the results are good, of course |
| 20:53 | <jamesr> | AryehGregor: in chrome try the timeline panel in the inspector |
| 20:53 | <AryehGregor> | How will the timeline help, without actual profiling? |
| 20:54 | <AryehGregor> | Changing to not append stuff to the DOM makes it take like 4 minutes in Chrome and 2 minutes in Firefox on my desktop instead of like 20 minutes . . . |
| 20:54 | <jamesr> | ah, well there's also a profiler |
| 20:54 | AryehGregor | might not need this VM thingie after all |
| 20:54 | <jamesr> | yeah always always batch |
| 20:57 | <AryehGregor> | I didn't think it would be an issue given how much work I was doing per test . . . |
| 21:23 | <AryehGregor> | It looks like there are some details of Unicode handling that don't work quite the same in Gecko/WebKit between Windows and Linux. |
| 21:23 | <AryehGregor> | Annoying. |
| 21:33 | <AryehGregor> | Turns out the Windows behavior is more correct, as it happens. |
| 21:35 | <AryehGregor> | The result of today's work: https://dvcs.w3.org/hg/editing/rev/8d721ff655d2 |
| 21:35 | <AryehGregor> | On AWS, how do I check what sort of bill I've run up so far? |
| 21:36 | <AryehGregor> | I don't seem to see that anywhere. |
| 21:36 | <jgraham> | AryehGregor: I wouldn't be surprised if people implemented non-N^2 string append in js for some benchmark and didn't do the same in DOM due to lack of benchmark |
| 21:37 | <AryehGregor> | jgraham, sure, but making it not O(n^2) is surely why appendData() and such actually exist . . . oh well. |
| 21:37 | <Philip`> | AryehGregor: I think there's something on your account information page that links to billing stuff |
| 21:37 | <jgraham> | AryehGregor: No reason it has to be n^2 in any case |
| 21:37 | <Philip`> | but I suppose that's a bit vague |
| 21:38 | <Philip`> | (Also, I don't know how frequently it's updated) |
| 21:38 | <AryehGregor> | jgraham, I guess not, if you treat the property magically. |
| 21:39 | <AryehGregor> | In theory foo.textContent += bar would be O(n^2), since it means foo.textContent = foo.textContent + bar, so you'd have to read textContent every time in full. But you could cheat on that easily enough. |
| 21:39 | <AryehGregor> | It'd have to be a special case, though. |
| 21:39 | <AryehGregor> | appendData() isn't a special case. |
| 21:39 | <AryehGregor> | Well, whatever. |
| 21:39 | <AryehGregor> | What I did seems to work, so awesome. |
| 21:41 | <AryehGregor> | Philip`, wow, they try to make this difficult. |
| 21:42 | <zewt> | AryehGregor: well, depending on the refcounting mechanism, i'd expect... |
| 21:42 | <zewt> | rather, GC mechanism |
| 21:43 | <zewt> | for example, CPython can implement nonexponential string appending, because it's refcounted and can tell if anything else has a reference to a string--if not it can avoid making the copy |
| 21:43 | <zewt> | other implementations use other GC methods which don't give that information, so they can't make that optimization |
| 21:44 | <zewt> | AryehGregor: billing is "account activity" |
| 21:45 | <zewt> | looks like they reorganized the console recently--have to click your name in the top-right corner (for some reason) |
| 22:00 | <gsnedders> | zewt: Most JS engines have string implementations that allow a string to be formed of a list of strings. |
| 22:01 | <gsnedders> | zewt: I thing CPython does something more complex. |
| 22:03 | <jamesr_> | JSC and V8 have rope support, i believe |
| 22:03 | <jamesr_> | fuzzy on the details |
| 22:03 | <jamesr_> | but that isn't necessarily useful in the DOM implementation |
| 22:03 | <gsnedders> | jamesr_: It won't be useful at all in the DOM impl |
| 22:03 | <gsnedders> | jamesr_: Carakan does too |
| 22:03 | <gsnedders> | jamesr_: And I'd be amazed if SpiderMonkey didn't |
| 22:05 | <zewt> | CPython just goes "if refcount == 1, append in-place", i think |
| 22:05 | <gsnedders> | jamesr_: Actually, no, we don't have ropes. We have something even ligher-weight that effectively is just lazyly evaluated concatenation |
| 22:05 | <jamesr_> | yeah, there are many ways to skin that cat |
| 22:05 | <jamesr_> | with the desired properties |
| 22:05 | <zewt> | poor cats |
| 22:06 | <gsnedders> | Lazely evaluating it keeps many of the O(1) properties of strings, at least |
| 22:17 | Philip` | wonders how many ways gsnedders can misspell "lazily" |
| 22:19 | <gsnedders> | Philip`: Lots. |
| 22:42 | <roc> | AryehGregor: what details? Gecko doesn't use much from the underlying platform so I'd be surprised |
| 23:09 | <zewt> | https://encrypted.google.com/ ... why is google drawing an arrow pointing at an empty space? heh |
| 23:10 | <zewt> | "dude, the corner of your monitor is pretty cool"? |
| 23:11 | <jamesr> | lawl |
| 23:11 | <jamesr> | i think that's where the google plus thingy shows up when you are logged in |
| 23:11 | <jamesr> | zewt: try google.com |
| 23:12 | <jamesr> | there's something there in the non-https version of that page |
| 23:12 | <zewt> | but i don't use google.com, i use encrypted.google.com |
| 23:12 | <zewt> | smells like poor testing practices for the front page of google's flagship product, heh :P |
| 23:13 | <zewt> | on google.com it points at "Web" (which is where you already are) |
| 23:14 | <zewt> | (and clicking it says "hey, you wasted your time clicking this since g+ is broken on gapps accounts") |
| 23:17 | <dglazkov> | zewt: :D |
| 23:19 | <zewt> | D: |
| 23:21 | <dglazkov> | don't be sad. sadness leads to poor digestion. poor digestion leads to fear. Fear leads to anger. Anger leads to hate. Hate leads to suffering. And more poor digestion. |
| 23:22 | <zewt> | i'd be less sad if google's main search page would stop intentionally distracting users from what they came to search for D: |
| 23:23 | <zewt> | i've never ran gdb and had it load pacman |
| 23:26 | <dglazkov> | zewt: I filed a bug -- does that make you feel a bit better? :) |
| 23:27 | <zewt> | on gdb asking for pacman? D: |
| 23:27 | <dglazkov> | sh> gdb |
| 23:27 | <zewt> | the frontpage stuff usually only lasts a couple days so it's sort of a self-expiring bug :) |
| 23:27 | <dglazkov> | > call pacman |
| 23:28 | <dglazkov> | ChrisWilson!!! |
| 23:31 | <ChrisWilson> | dglazkov! |
| 23:33 | <dglazkov> | ChrisWilson: I honestly don't know why I did that. |
| 23:47 | <ChrisWilson> | :P |