User:Enkidu Linden/Office Hours/2010-05-13

List of Attendees

• Enkidu Linden
• ImSky Amiot
• Listing disappears
• Morgaine DinovaMorgaine
• Morgaine Dinova
• NaoWut DigfootNaoWut
• NaoWut Digfoot
• New Search
• Xugu MadisonXugu
• Xugu Madison

Transcript

[11:02]	Xugu Madison:	Mixed. Ignoring feature requests, which is helping keep the backlog short *shuffles*
[11:02]	Xugu Madison:	hey Enkidu!
[11:02]	Enkidu Linden:	hey there!
[11:03]	Morgaine Dinova:	Hi Enkidu :-)
[11:03]	Enkidu Linden:	how is everyone?
[11:04]	Morgaine Dinova:	A week of nothing, don't think it'll make the history books :P
[11:04]	Xugu Madison:	Not bad, not bad. How's you Enki?
[11:04]	Enkidu Linden:	i could use about 10 hours of sleep
[11:04]	Morgaine Dinova:	Me too!
[11:04]	Morgaine Dinova:	I vote we defer Thursday for 10h
[11:05]	Enkidu Linden:	we could take a nap now
[11:05]	Enkidu Linden:	that'd leave us at -9
[11:05]	Morgaine Dinova:	I'm way ahead of you :P
[11:05]	Enkidu Linden:	haha
[11:05]	Xugu MadisonXugu	Madison gently falls forwards, using Enkidu as a jellyfish pillow "Works for me"
[11:06]	Morgaine Dinova:	Hmmm, lots of scientific progress on physiology+anatomy, has any of it hinted at a way of reducing/eliminating sleep requirements?
[11:07]	Xugu Madison:	Morgaine; yes, although I think mostly people edge around that stuff
[11:07]	Enkidu Linden:	there's the uberman sleep schedule
[11:07]	Xugu Madison:	polyphasic sleep is apparently fairly workable, at least for 2-3 years
[11:07]	Morgaine Dinova:	What's that Xu?
[11:07]	Enkidu Linden:	http://en.wikipedia.org/wiki/Polyphasic_sleep #Uberman.27s_sleep_schedule
[11:08]	Morgaine DinovaMorgaine	Dinova looks
[11:08]	Xugu Madison:	What Enki said :)
[11:08]	Enkidu Linden:	i wanted to try it but i was already married
[11:08]	Xugu Madison:	Careless Enkidu, very careless. Wouldn't find me getting married, oh no..
[11:08]	NaoWut DigfootNaoWut	Digfoot waves
[11:09]	Enkidu Linden:	i'm in dangerous territory
[11:09]	Enkidu Linden:	hello naowut
[11:09]	Xugu Madison:	hi NaoWut!
[11:09]	NaoWut Digfoot:	hiya enkidu, xugu, blookitty :)
[11:09]	NaoWut Digfoot:	is there an agenda for todays OH?
[11:09]	Morgaine Dinova:	Hi Wut :-)
[11:09]	Morgaine Dinova:	Yep, we're discussing scalable non-sleeping web architectures
[11:10]	Xugu Madison:	Maybe we can distribute sleeping? Can I offshore my sleep to India?
[11:10]	NaoWut Digfoot:	totally out of my field here, but sounds interesting :)
[11:10]	NaoWut Digfoot:	Xugu =P
[11:10]	Enkidu Linden:	mmm distributed restful resting
[11:11]	Morgaine Dinova:	Ah! So that's why I like RESTful interfaces
[11:11]	NaoWut Digfoot:	hehe
[11:11]	Xugu Madison:	I actually go through insomnia during summer. Except to be extra special, I wake up early. I've been on 6-ish hours sleep/night for about a month now...
[11:12]	Enkidu Linden:	as for the agenda, i thought it might be interesting to discuss something i saw on ajaxian this morning
[11:12]	Enkidu Linden:	http://eligrey.com/blog/post/localization-in-javascript
[11:13]	Xugu Madison:	Can't we just teach everyone esperanto? :)
[11:14]	Enkidu Linden:	perhaps then we could all just get along
[11:14]	Morgaine Dinova:	It's not in the interest of those in power for people to get along
[11:15]	Xugu Madison:	So looking at using this to localise the websites, Enkidu?
[11:15]	Enkidu Linden:	yeah, too true. harmony drowns out the single note
[11:15]	Enkidu Linden:	well, actually i wouldn't use this
[11:16]	Enkidu Linden:	that would be far too easy
[11:16]	Enkidu Linden:	joke!
[11:16]	Morgaine Dinova:	That link suggests that no reduction to the 8h sleep is achieved by changing the duration of individual sleep periods.
[11:17]	Enkidu Linden:	the actual problem of client-side localization really boils down to managing data transfers
[11:17]	Enkidu Linden:	i mean, it boils down to that for me because my principle concern is performance
[11:18]	Morgaine Dinova:	How does localization affect performance?
[11:18]	Enkidu Linden:	i'm so glad you asked!
[11:18]	Morgaine Dinova:	:P
[11:19]	Enkidu Linden:	now if you'll just climb into this box while i get the saw..
[11:19]	Enkidu Linden:	most of the time when we localize we do it on the server side
[11:20]	Morgaine Dinova:	Blue Kitties are higher dimensional, that'll have to be a special saw
[11:20]	Enkidu Linden:	we get a request and encoded in the request somehow is the locale, and the server can choose a set of localized strings and interpolate
[11:21]	Morgaine Dinova:	So what causes a perf hit? Doesn't seem like a big issue at first glance
[11:21]	Enkidu Linden:	that's true as long as you're willing to make a request, but many "web 2.0" applications don't always do that
[11:22]	Morgaine Dinova:	Request from what to what?
[11:22]	Enkidu Linden:	from browser to web server
[11:22]	Enkidu Linden:	let me paint some context
[11:22]	Morgaine Dinova:	kk
[11:23]	Enkidu Linden:	facebook, for example, goes to great lengths to avoid full page loads
[11:23]	Enkidu Linden:	they do this by getting more of the application (bunch of javascript) into the browser in the first load
[11:24]	Enkidu Linden:	then when you click on something, like "Older Posts" or whatever that thing is called at the bottom of whatever they call it these days..
[11:24]	Enkidu Linden:	it can get the data through a service call and display it
[11:24]	Xugu MadisonXugu	Madison quite fancies the distributed Facebook-like apps some people are trying...
[11:25]	Enkidu Linden:	now, the service data could be easiy localized, but there's application strings that go along with displaying that data
[11:26]	Enkidu Linden:	so the browser has to already have those strings or it has to get them
[11:27]	Enkidu Linden:	which is a choice between preloading all the app's strings, most of which won't even be used in any given page view, or requesting those strings as needed
[11:27]	Enkidu Linden:	and there's a lot of room in-between for some hybrid appraoch
[11:29]	Xugu Madison:	if you can serve the application strings from a static file, and ensure it's got an Etag header or similar, surely caching can handle most of the load?
[11:30]	Morgaine Dinova:	Well if you want performance, just get rid of all the Javascript altogether, and localize by preformatting cached templates for each language.
[11:30]	Morgaine Dinova:	JS is just a disease in most cases, certainly not necessary
[11:31]	Enkidu Linden:	caching is a great mitigator, but we sometimes think that because things are cached, they're fast
[11:31]	Enkidu Linden:	"when things are cached, they're fast" is correct
[11:32]	Morgaine Dinova:	Storage costs nothing, CPU is as costly as always. The clever approach is to cache.
[11:32]	Enkidu Linden:	rather than "i have enabled caching, so things are fast"
[11:32]	Enkidu Linden:	let me restate slightly
[11:33]	Xugu MadisonXugu	Madison nods "However, I'd assume you see a lot of repeat traffic, so cache misses are few, and initial loading isn't a major issue?"
[11:33]	Enkidu Linden:	actually let me back up a bit
[11:33]	Xugu MadisonXugu	Madison puts Enki into reverse
[11:34]	Enkidu Linden:	caching is wonderful but it doesn't address the worst case
[11:34]	Xugu MadisonXugu	Madison nods "However, is the worst case tolerable if caching handles the common case?"
[11:34]	Enkidu Linden:	there you go
[11:35]	Xugu MadisonXugu	Madison tends to assume it is :)
[11:36]	Morgaine Dinova:	Don't assume, measure
[11:36]	Enkidu Linden:	so there's a little web performance nugget of wisdom: cache! but verify.
[11:37]	Xugu MadisonXugu	Madison only has about 10k users, and can't understand why people worry so much *grin* *duck*
[11:37]	Morgaine Dinova:	To go with the other nugget of "What you don't measure you don't really know, even when you think you do." :-)
[11:38]	Enkidu Linden:	yes, metrics are really, really important
[11:39]	Enkidu Linden:	as it happens we have a small, highly cacheable public service that performs rather poorly on cache misses
[11:41]	Enkidu Linden:	http://secondlife.com/community/classifieds/?lang=en-US
[11:41]	Enkidu Linden:	the thumbnail image generator for that page is pretty slow
[11:41]	NaoWut DigfootNaoWut	Digfoot returns from afkness and reads backlog
[11:42]	Enkidu Linden:	it'll load fast for you because they're all on the front page of classifieds and likely all cached
[11:43]	Enkidu Linden:	http://secondlife.com/community/classifieds/index.php?search_maturity=PG&limit=15&offset=150
[11:43]	Enkidu Linden:	if you change the offset there to something random and > 200 you should see the images loading rather slowly
[11:43]	NaoWut Digfoot:	ow. yup.
[11:45]	Enkidu Linden:	haha, massive digression
[11:45]	Enkidu Linden:	i love caching though
[11:46]	NaoWut Digfoot:	Morg: while i agree js is a massive disease in the wrong hands, done properly it can do wonders
[11:46]	Enkidu Linden:	morgaine, your suggestion to load html fragments instead of data rendered by js is also good in some situations
[11:46]	Enkidu Linden:	we do that on the sl.com dashboard
[11:48]	Enkidu Linden:	you could also include app strings in the initial load that are likely to be used based on some analysis, and just request the rare stuff as needed
[11:48]	Morgaine Dinova:	Well JS is often used merely because it's "the thing to use", rather than because it makes sense.
[11:49]	NaoWut Digfoot:	morg, yup. and there it goes bad.
[11:49]	NaoWut Digfoot:	and yeah, i like that idea enkidu
[11:50]	NaoWut Digfoot:	morg, also it's the common case :)
[11:50]	Morgaine Dinova:	Reminds me of a C program I once saw containing nothing but system("<text>") calls. You know, sometimes people just use the wrong solutions.
[11:51]	Enkidu Linden:	yikes
[11:51]	Enkidu Linden:	i've seen that approach as well
[11:52]	Enkidu Linden:	i always think, "what if instead of trying to confuse me you just write this as a shell script?"
[11:53]	NaoWut Digfoot:	hehehe, oh boy. yes some people are prone to that sort of programming kludgery
[11:54]	Enkidu Linden:	so hey, maybe next week we should talk about caching
[11:54]	ImSky Amiot:	or maybe if your talking servers and programming somthing to do with the search faults
[11:54]	NaoWut Digfoot:	with caching metrics too! :)
[11:55]	ImSky Amiot:	metircs aren't there enough false figures what the resdients want is a working search system
[11:55]	NaoWut Digfoot:	i was referring to web caching, ImSky
[11:56]	ImSky Amiot:	k
[11:56]	Enkidu Linden:	i think you can expect to see improvements to search
[11:56]	Enkidu Linden:	i'm not on the search team though
[11:56]	ImSky Amiot:	we have seen over 50% of places removed in the last two weeks
[11:57]	ImSky Amiot:	and no one listening jsut saying its the new system
[11:57]	ImSky Amiot:	i have 5 dead shops and my rl living is here and i'm only one of hundreds in the same situation
[11:57]	Enkidu Linden:	have you opened a pjira?
[11:58]	NaoWut Digfoot:	I'm sorry to hear that ImSky
[11:58]	Enkidu Linden:	yeah, me too
[11:58]	ImSky Amiot:	yep and thats a waste of time, as was a live chat, tickets, tryingot contact the search team
[11:58]	Enkidu Linden:	can you give me the pjira link?
[11:59]	ImSky Amiot:	there are several
[11:59]	ImSky Amiot:	one sec
[11:59]	Enkidu Linden:	that's ok, you can give me all of them
[11:59]	Enkidu Linden:	i can't fix them myself but i can talk to some people
[11:59]	Morgaine Dinova:	There's a Search team? Unlikely. If there was, they were clearly fired in order to bring Viewer2.0 out searchless.
[12:00]	ImSky Amiot:	http://jira.secondlife.com/browse/SVC-5789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=186179 #action_186179
[12:00]	New Search	is not working correctly at all
[12:00]	NaoWut Digfoot:	Morg: shush :P
[12:00]	ImSky Amiot:	there is liana linden is head of it and sea linden is also a member
[12:00]	Morgaine Dinova:	No no, I'm serious. Search worked, and it's hard to muck up search accidentally. It was deliberately sabotaged for V2.0. The question is why.
[12:01]	ImSky Amiot:	answer is obvious
[12:01]	ImSky Amiot:	new viewer was bruaght out for a better web interface for street
[12:01]	ImSky Amiot:	more commsions that way
[12:02]	ImSky Amiot:	new search is so mucked up that it psuhes all to street instead
[12:02]	ImSky Amiot:	have you tried searching for anything recently
[12:02]	ImSky Amiot:	its a joke
[12:02]	Morgaine Dinova:	That's a possible reason. World destruction is OK, as long as profits from xtreeet maximized. I don't think there's evidence to support that reason. However, search was definitely sabotaged. It could never have passed testing.
[12:03]	Morgaine Dinova:	Not even a single test could have suceeded, let alone the large suite that must exist.
[12:03]	Morgaine Dinova:	So it was on purpose.
[12:03]	ImSky Amiot:	then how come lotws of palces simply have lost there listings
[12:03]	NaoWut Digfoot:	hmm, failing on the server side? or viewer-2?
[12:03]	ImSky Amiot:	maturtiy ratings were a known bug in it and its been owned up to
[12:03]	ImSky Amiot:	so tetign wasnt' too good
[12:04]	ImSky Amiot:	any category filters out places
[12:04]	Enkidu Linden:	sorry, i was reading the jira
[12:04]	ImSky Amiot:	whats on places first page is often not in any category at all
[12:04]	Morgaine Dinova:	Yeah but the loss of search isn't a mere bug. That simply would not have passed a single test. That was deliberate. LL is a competent outfit, and total loss of search could not have passed QA by accident.
[12:05]	ImSky Amiot:	clubs are taking first places on keywords for items "beach house" frist page is full of clubs
[12:05]	ImSky Amiot:	so your satying we are stuck with it as it was an act done on purpose
[12:06]	Morgaine Dinova:	Well I can't see how it could be accidental. It would mean that the whole QA division is on holiday.
[12:06]	ImSky Amiot:	is there such a thing?
[12:06]	ImSky Amiot:	we have seen so many bugs recently
[12:06]	NaoWut Digfoot:	i couldn't make that judgement call. but yes, i see results are much smaller than i was expecting
[12:07]	Morgaine Dinova:	Yep, there is a QA division, and fairly big too. Enus is in charge, he used to hold OHs, very informative.
[12:07]	ImSky Amiot:	anyone knwo how or what needs to be doen to gwet somone to listen t this
[12:07]	Enkidu Linden:	wow, i had no idea this was such a problem
[12:08]	NaoWut Digfoot:	hum
[12:08]	NaoWut Digfoot:	http://jira.secondlife.com/browse/VWR-18792
[12:08]	Listing disappears	in "search" when more than one maturity rating is check marked
[12:08]	NaoWut Digfoot:	that
[12:08]	NaoWut Digfoot:	is curious behaviour
[12:08]	ImSky Amiot:	itsd a huge problem, i'm hearing it from build, sculpt, estate manger groups
[12:09]	NaoWut Digfoot:	do they get the same issues in snowglobe 2.0? that would be interesting to check
[12:09]	ImSky Amiot:	worse in viewer 2
[12:09]	NaoWut Digfoot:	:/ seems strange that they'd mess with that
[12:10]	ImSky Amiot:	in 1.23 and emerlad you have the drop list and can some if you drop to places. in viewer 2 you dont' have that option
[12:10]	Morgaine Dinova:	No idea. I use Imprudence. The 1.x era viewers work.
[12:10]	Enkidu Linden:	hmm
[12:11]	Morgaine Dinova:	Enki: the whole thing is kinda surreal. ;-)
[12:11]	NaoWut Digfoot:	horrific braindamage in search fail. viewer 2.0 search should surely be like snowglobe 2.0?
[12:11]	NaoWut Digfoot:	xD
[12:11]	ImSky Amiot:	what weneed is to fidn out how we get our shops parcels back inotsearch, i've meesed with parcel descriptions, prim names, prodcut names till i'm blue in the face
[12:12]	ImSky Amiot:	and worst of all whats turning up in search seems to be irrelevent
[12:12]	Enkidu Linden:	for what it's worth i don't think there was any intent to diminish the quality of search. quite the opposite!
[12:12]	ImSky Amiot:	whatever the intention a fix is needed
[12:12]	NaoWut Digfoot:	no,the consipiracy theory would be a pretty BIG own goal
[12:13]	Morgaine Dinova:	I think the explanation based on "deliberate" is actually more generous.
[12:13]	NaoWut Digfoot:	and by that. i mean bigger than usual
[12:14]	Enkidu Linden:	guys, we're over time. i will be posting a transcript as usually and will definitely make sure some other lindens see it
[12:14]	NaoWut Digfoot:	who would stand to gain? you spend i have no idea how much developing a new viewer, outsourcing everything to graphics designers, but then deliberatey sabotage search in the client side?
[12:14]	Xugu Madison:	Going to have to scamper, take care all!
[12:14]	ImSky Amiot:	its a nice peice of work fro them a bugy search pushing more to street but non shops are gettign hit with things they dont' want and residents arent' findign what they want.
[12:14]	ImSky Amiot:	take care
[12:14]	NaoWut DigfootNaoWut	Digfoot nods, thanks for the meeting enkidu, next time, caching :)
[12:15]	Morgaine Dinova:	But theories aside, there's no doubt it's all very bizarre / surreal.
[12:15]	Xugu Madison:	Thanks for hosting Enkidu!
[12:15]	Morgaine Dinova:	Cool, thanks Enki :-)
[12:15]	NaoWut Digfoot:	right. i'm just having a poke diffing the message_template.msg to see if there were new fancy but broken search packets
[12:15]	couch sit	1: llStopAnimation: Script trying to stop animations but agent not found
[12:16]	ImSky Amiot:	the biggest problem seems to the filtering of options
[12:16]	ImSky Amiot:	and i suspect the prob is there
[12:16]	Enkidu Linden:	thanks for coming, everyone!
[12:16]	Enkidu Linden:	have a great week!
[12:17]	NaoWut Digfoot:	and you Enkidu :)
[12:17]	Morgaine Dinova:	You too Enki :-)

Generated with SLog Wikifier