User:Zero Linden/Office Hours/2008 October 07

From Second Life Wiki
Jump to: navigation, search
  • [12:59] Saijanai Kuhn:  : But is it *clean* coal, Zha?
  • [13:00] Zha Ewry: laughs
  • [13:00] Whump Linden: burns clean burning sim coal!
  • [13:00] Zha Ewry: Nah. Gorean. What do you expect in SL?
  • [13:00] Saijanai Kuhn:  : Gorean...
  • [13:00] Tao Takashi: Hello ppl of Zero's OH
  • [13:00] Saijanai Kuhn: shudders and goes back to sleep dreaming of scantily clad women with whips and swords. Oh great.
  • [13:00] FWord Utorid: ok, shh... the toaster is here
  • [13:01] Eshi Otawara: greetings
  • [13:01] Meadhbh Oh: um... Sai... how did you figure out what my halloween costume was...
  • [13:01] Morgaine Dinova: Hiya Whump. Is Zero here today?
  • [13:01] Whump Linden: Hey, so Zero appologizes. He's kinda toasty today.
  • [13:01] Morgaine Dinova: Aha
  • [13:02] Whump Linden: So I'm subbing for him.
  • [13:02] Whump Linden: So, topics?
  • [13:02] Thoys Pan: havent heard any word yet =d
  • [13:02] Thoys Pan: hello Latif =)
  • [13:02] Latif Khalifa: hi all :)
  • [13:02] Whump Linden: Hi Latif.
  • [13:03] Whump Linden: Zero is unavailable this afternoon, so I'm subbing.
  • [13:03] Dahlia Trimble: Hi Latif :)
  • [13:03] Saijanai Kuhn: mutters something about legacy group IM and goes back to sleep. Darned whips
  • [13:03] Morgaine Dinova: What's your speciality, Whump? We could stick with yours, since you're hosting, ideal opportunity ;-)))
  • [13:03] Zha Ewry: I think we should hold naked UUIDs, until Zero makes it in
  • [13:03] Saijanai Kuhn:  : UUIDs with swords and whips. Oh great...
  • [13:03] Tao Takashi: so let's talk about dressed UUIDs then :)
  • [13:03] Whump Linden: Well, if folks have questions related to the Open Grid Public Beta.
  • [13:04] FWord Utorid: that sounds interesting
  • [13:04] Tao Takashi: naked is not allowed in SL anyway
  • [13:04] Thoys Pan: it isnt Tao?
  • [13:04] Whump Linden: Or Pyogp since Sai and Saij are here.
  • [13:04] Thoys Pan: oops
  • [13:04] Saijanai Kuhn:  : Tao and Saij..
  • [13:04] Tao Takashi: the Sai brothers :)
  • [13:04] FWord Utorid: a modern day cheech and chong?
  • [13:04] Whump Linden: er, I'm typing double.
  • [13:04] Morgaine Dinova: Tao and Sai are always here, more reliable than the sun ;-)
  • [13:04] Thoys Pan: im walkin behind, when will this beta be?
  • [13:05] Meadhbh Oh: hmm... since when is naked not alloewd in SL?
  • [13:05] Whump Linden: The beta is currently underway.
  • [13:05] FWord Utorid: yeah, i would like to hear more about the beta myself, and what's entailed in it
  • [13:05] Tao Takashi: well, I am a bit out of pyogp at the moment due to vacation and work
  • [13:05] lyndell Aleixandre: im quite intrested in that beta
  • [13:05] Tao Takashi: I always hear about "broadly offensive" and as it's not defined I am on the safe side to say it's not allowed ;-)
  • [13:05] FWord Utorid: is this essentially the gridnauts SL > OSGrid endeavor, or are there more terms to learn?
  • [13:05] Whump Linden: Okay, so I think most people here are familiar with the AWG, correct?
  • [13:05] Tao Takashi: but that's a different topic for a different OH
  • [13:06] Eshi Otawara: what's awg? i suck at abbreviations
  • [13:06] Tao Takashi: I personally wonder about the new viewer design M was talking about but I guess that won't be a topic
  • [13:06] FWord Utorid: eshi, architecture working group
  • [13:06] Eshi Otawara: oh that.
  • [13:06] Eshi Otawara:  :)
  • [13:06] lyndell Aleixandre: AWG is why we are here in the first place is it not?
  • [13:06] Meadhbh Oh: M was talking about a new browser design?
  • [13:06] Meadhbh Oh: i'm out of it
  • [13:07] Latif Khalifa: Tao, I am very afraid of "n00b friendly" viewer ;)
  • [13:07] Whump Linden: So the Architecture Working Group has been developing, along with Zero and other developers at the Lab, a set of protcols for inter-virtual-world operablity.
  • [13:07] Tao Takashi: well, hiistorically this OH wasn't about AWG but more about SL architecture and inner workings in general
  • [13:07] FWord Utorid: this is zero lindens office hours, hosted by lump whinden, er, whump linden, not an AWG meeting.
  • [13:07] FWord Utorid: but it's going to turn into an AWG meeting if whump says so
  • [13:07] Tao Takashi: I want to add that it's not yet finished
  • [13:07] Tao Takashi: but of course we are close to finishing it, it's all in our heads and we just need to write it down, isn't it? :)
  • [13:08] Whump Linden: So the first set of protocols have been developed around teleport.
  • [13:08] Teravus Ousley: heh
  • [13:08] Eshi Otawara: noob viewer- easy- teleport to a sex place button, undress button, hump button and beg for money button.
  • [13:08] Meadhbh Oh: and login
  • [13:08] Tao Takashi: we might end up with 20 versions but hey, competition is not that bad ;-)
  • [13:08] Dahlia Trimble: Inventory!!!! :)
  • [13:08] Teravus Ousley: If there are 150 'grid' messages.. and we've done 7.. we've got a long way to go.
  • [13:08] Whump Linden: Yes, just need to find the cable that will go from our heads to the printer.
  • [13:08] Dahlia Trimble: oops that slipped :/
  • [13:08] Meadhbh Oh: yay! inventory! that's a good topic
  • [13:09] Latif Khalifa: Whump, has there been any work on implementation of anything beyond login and teleport?
  • [13:09] Whump Linden: Not at the moment.
  • [13:09] Whump Linden: If you have suggestions, please add them to https://jira.secondlife.com/browse/SVC-3156
  • [13:09] Strawberry Fride: I'd be happy if I could transfer inventory from one alt to another, let alone cross grid
  • [13:09] lyndell Aleixandre: so what are the next most critcal messages needed to be worked on?
  • [13:10] Rex Cronon: hello everybody
  • [13:10] Tao Takashi: not sure if there is anything more critical than the other, maybe more or less complex but in the end we need it all
  • [13:10] Dahlia Trimble: Hi Rex :)
  • [13:10] lyndell Aleixandre: i think we should start with inventory
  • [13:10] Whump Linden: If you don't have a headset on, Teravus mentions: inventory, trust, instant messaging.
  • [13:10] Rex Cronon: hi
  • [13:10] Saijanai Kuhn:  : group IM rototyping covers all the bases ;-)
  • [13:10] Dahlia Trimble: assets
  • [13:10] FWord Utorid: teravus, you gotta start typing dude
  • [13:10] Dahlia Trimble: oops slipped again :(
  • [13:11] Teravus Ousley: hehe
  • [13:11] Saijanai Kuhn:  : prototyping. rotoscoping?
  • [13:11] FWord Utorid: stereotyping, sai
  • [13:11] Saijanai Kuhn:  : bingo
  • [13:11] Latif Khalifa: yeah, how will assets work. does every asset need to have UUID + domain now?
  • [13:11] Tao Takashi: well, to make a useable world it might not be inventory or group IM but chat
  • [13:11] Meadhbh Oh: i like the concept of rotoscoping software... you just draw around the bits you don't want to see in the finished product
  • [13:11] Saijanai Kuhn:  : though inworld chat was still working in OGP
  • [13:11] Whump Linden: and if people are sick to death about talking about ogp, since this is a general OO we can hit other topics.
  • [13:12] lyndell Aleixandre: mmmm uuid's are tricky now that you have to deal with seerate asset servers
  • [13:12] lyndell Aleixandre: *seperate
  • [13:12] Tao Takashi: well, in-world chat is working as it's not changed from the original protocol but it's not yet defined in OGP
  • [13:12] FWord Utorid: seprit
  • [13:12] Meadhbh Oh: uh oh... the question from Lati begs a discussion re: nekkid UUIDs
  • [13:12] Whump Linden: Sorry, Zha, we're going down the UUID path again.
  • [13:12] Tao Takashi: nor I guess is what happens after you teleported somewhere
  • [13:12] Latif Khalifa: yeah, I guess you'd need asset UUID + some sort of dmain
  • [13:12] Dahlia Trimble: UUIDs are so boring :(
  • [13:12] Teravus Ousley: domain + uuid +++;
  • [13:12] FWord Utorid: i agree, dahlia.
  • [13:12] lyndell Aleixandre: last meeting we went down the instant messaging path
  • [13:12] Tao Takashi: yeah, URLs are the way to go :)
  • [13:12] Meadhbh Oh: but wouldn't the asset come from the simulator tht's getting it from the agent domain?
  • [13:13] Meadhbh Oh: (leading question)
  • [13:13] FWord Utorid: but the things we could talk about that aren't boring, sai would find objectionable, and we have to be considerate of him for some reason
  • [13:13] Tao Takashi: I would think you copy (or move depending on perms) an object from your AD to the RD when you rez it
  • [13:13] Latif Khalifa: Well lets say I teleport to a sim. How would viewers of the people around me know where to download my skin from?
  • [13:13] Zha Ewry: looks up and sighs
  • [13:13] Zha Ewry: No Naked UUIDs without context
  • [13:13] Tao Takashi: I think now we have 20 topics ;-)
  • [13:14] Saijanai Kuhn:  : https://wiki.secondlife.com/wiki/AW_Groupies/Chat_Logs/AWGroupies-2008-10-07
  • [13:14] Dahlia Trimble: UUIDs must be clothed in context
  • [13:14] lyndell Aleixandre: lets stick to one hmm?
  • [13:14] Whump Linden: Yes, stick with UUIDs for the moment.
  • [13:14] Rex Cronon: u put a stamp on your skin, that says: to download me go to yahoo.com and pay 10L$:)
  • [13:14] Morgaine Dinova: Make a short list Whump, as we're already 1/4 way through ;-)
  • [13:14] FWord Utorid: we *never* stick to one topic. usually, the topic is about what the topic is.
  • [13:14] Meadhbh Oh: Terevus makes me sad to think that Don LaFontaine recently passed
  • [13:14] Tao Takashi: I would like to hear more about the use of naked UUIDs but more so on a mailing list
  • [13:14] lyndell Aleixandre: and keep offtopic chatter to the minimum this time :)
  • [13:14] Tammy Nowotny: my skin is a freebie :-)
  • [13:15] Teravus Ousley: terrible loss
  • [13:15] Tao Takashi: and my point is still that we should keep concepts in OGP down and just use URLs instead of UUIDs
  • [13:15] Rex Cronon: than is zero:)
  • [13:15] lyndell Aleixandre: thats slow though isnt it?
  • [13:15] Tao Takashi: and keep this question on the implementation level
  • [13:16] Meadhbh Oh: well.. caps are definitely URLs
  • [13:16] Meadhbh Oh: but they frequently have UUIDs in them
  • [13:16] Tao Takashi: and that's an implementation issue as well
  • [13:16] lyndell Aleixandre: we cant solve all the problems by just making them url's
  • [13:16] FWord Utorid: a ton of viewer to simulator calls, such as sitting, are uuids
  • [13:16] Tao Takashi: I still wonder: what's the problem?
  • [13:17] Whump Linden: lyndell, what problems are left over if we make them URLs?
  • [13:17] Zha Ewry: We can. Two questoins come to mind, Hw much lnger are they, and how much does it costs to turn them into easy to cache and lookup tokens client side?
  • [13:17] Meadhbh Oh: FWord... but they don't need to be in the future
  • [13:17] FWord Utorid: and in the long run, everything should be an URL
  • [13:17] Saijanai Kuhn:  : mornign (afternoon?) Other Fearless Leader
  • [13:17] Morgaine Dinova: Well the way Zha described UUIDs as a session shortcut to URIs between peers, it implies that both URIs and UUID must be valid entities in OGP, if OGP is to be the only protocol between peers.
  • [13:17] Tao Takashi: well, viewer calls are right now using UUIDs but we are talking about OGP and we can define them to be whatever we want them to be I assume
  • [13:17] Tao Takashi: some might make more sense though than others
  • [13:17] FWord Utorid: Meadhbh, agreed, just pointing out the kind of behavior that will break if there's a shift from UUID to URL
  • [13:17] Meadhbh Oh: you could define a service location for each object that's viewable
  • [13:17] Meadhbh Oh: and one of the interfaces could be "sit_here"
  • [13:18] lyndell Aleixandre: its not that they may or may not be url's whump .. but i tend to think that overusing XML or HTTP for everything will leave a large protocol overhead in exchange for ease of implimentation
  • [13:18] FWord Utorid: URL's in conjunction with most functionality in the VW makes it more accessible, more easily programmable, and will invite new viewer paradigms faster.
  • [13:18] Tao Takashi: well, you need some host to open a connection to anyway. so your URL could be as short as <UUID>
  • [13:19] Tao Takashi: but the UUID part would be implementation specific and it's not much longer than just a UUID
  • [13:19] Saijanai Kuhn:  : or even [1]
  • [13:19] FWord Utorid: XML and HTTP are overused because they are proven standards, and the majority of the internet today is based on those protocols and structures
  • [13:19] FWord Utorid: there's good reason for that, it's a good mechanism for deployment
  • [13:19] lyndell Aleixandre: so the debate begins: speed vs interoperability
  • [13:19] FWord Utorid: simulator://sit_here/objectid
  • [13:19] Latif Khalifa: So in theory, the viewer would have to connect to 20 asset servers to fetch the textures needed to render a scene?
  • [13:20] Morgaine Dinova: Nah, for speed just choose JSON over XML, both should be available at the REST interface.
  • [13:20] lyndell Aleixandre: that mechanism in my opinion is very inefficient
  • [13:20] Tao Takashi: my point is also that we want this protocol to be implemented and as coders are also just humans and are thus lazy we should make it easy for them and not have more concepts than necessary IMHO.. given that they probably come with a web programming background it makes sense to reuse web standards I think. So that would be URLs (and JSON instead of LLSD btw ;-) )
  • [13:20] Rex Cronon: only 20? much more
  • [13:20] Meadhbh Oh: hmm... @morgaine... some languages parse XML about as fast as they parse JSON
  • [13:20] FWord Utorid: clicking on a prim == clicking a link on a webpage, it's a very good comparison
  • [13:21] Saijanai Kuhn: still isn't sure if the LLIDL can't be used in the same way JSON is
  • [13:21] Tao Takashi: Latif: The sim might be caching it for the viewer and return a URL to it's cache
  • [13:21] Dahlia Trimble: keyword: "some"
  • [13:21] Meadhbh Oh: @Sai... in which way?
  • [13:21] Tammy Nowotny: XML seems to have a bigger market share than JSON... actually waht is JSON, I am wondering?
  • [13:21] Meadhbh Oh: LLIDL is an abstract type system
  • [13:21] Meadhbh Oh: JSON is just one of the serializations for abstract types
  • [13:21] FWord Utorid: Tammy, JSON is JavaScript Object Notation.
  • [13:21] Morgaine Dinova: Meadhbh: those languages must have very poor JSONs then. :-) I've benchmarked a large number, and the variation in JSON quality is vast, 2+ orders of magnitude.
  • [13:21] lyndell Aleixandre: FW you know that small delay between clicking and the browser responding? imagine that only for every texture you see
  • [13:21] Whump Linden: lyndell: it's also a protocol that we hope to have broad impelemention (sp) so http and XML or JSON are well understood.
  • [13:22] Tammy Nowotny: ah OK, thanks FW :-)
  • [13:22] Latif Khalifa: Tao, that is the protocol questio I am asking. Right now the viewer only talks to a sim. Will the protocol specify that the viewer talks to asset servers directly, or will it be passed through the sim?
  • [13:22] FWord Utorid: it is pretty cool that it uses less text with JSON than XML, but there are many better parsers for XML than JSON
  • [13:22] FWord Utorid: there's a small delay between right clicking a prim and the menu that pops up too
  • [13:22] Saijanai Kuhn:  : the wiki's textual respresentation of sample LLIDL
  • [13:22] Meadhbh Oh: i think the biggest problem with XML is that standard bindings for some languages make it difficult to code parsers
  • [13:22] Meadhbh Oh: well.. not difficult..
  • [13:22] Meadhbh Oh: but harder than just using JSON
  • [13:23] Tao Takashi: Latif: good question.. but I personally lack the understanding of the details as it works now ;-) but I personally would also like to be able to specify URLs as textures for my objects instead of uploading them to some server
  • [13:23] Whump Linden: ah, like an optimized LLSD consumer
  • [13:23] Zha Ewry: One interesting point to keep in mind, is that the sim, probably has all the assets in hand, and you want that, so, you don't *have* to go to N asset servers when you wanrt things which
  • [13:23] FWord Utorid: once all of this work is done defining things, there will be a standard, and people will disagree with that standard, and there will be 1000 types of virtual worlds for people to argue about what's better.
  • [13:23] Zha Ewry: are simply sitting on the regoin simulator
  • [13:23] Tao Takashi: it might be also a question of some experiment how things would be different if textures would be loaded from 20 different places instead of just 1
  • [13:23] Morgaine Dinova: Well Zero's idea is to avoid representation wars and simply provide both XML and JSON serializations. May the best one win ;-)
  • [13:23] Meadhbh Oh: @Tao... wouldn't that imply using UUIDs for textures?
  • [13:24] Latif Khalifa: Tao, right now SIM is the ultimate proxy. The viewer talks only to the sim. And its the sim that passes through assets, messages, ims, whatnot
  • [13:24] Meadhbh Oh: so you would know what's cached locally?
  • [13:24] Rex Cronon: intiaally the sim, still has to download all the assets
  • [13:24] Zha Ewry: You can look at Cached URL as easily as UUIDs, if you crucnh them sanely
  • [13:24] Tao Takashi: Meadhbh: why? I can cache URL based assets right now
  • [13:24] Tao Takashi: just a key
  • [13:24] Meadhbh Oh: @Tao... good point
  • [13:24] Meadhbh Oh: yup
  • [13:24] Meadhbh Oh: just use the URI as an opaque id
  • [13:24] Tao Takashi: I might even use a varnish for it
  • [13:25] Zha Ewry: You may have the oddity of going to your local sim, asking for a URL which is not local, but HTTP is ok with that.
  • [13:25] Tao Takashi: I couldn't use varnish or squid if they would be UUIDs
  • [13:25] FWord Utorid: either way, it sounds to me like URL's will be much happier than UUIDs for most actions
  • [13:25] Meadhbh Oh: the only problem yo would encounter would be if you got the same texture from two URIs
  • [13:25] Meadhbh Oh: so
  • [13:25] Meadhbh Oh: just don't do that
  • [13:25] Tao Takashi: well, it would be again up to the implementation if it loads the textures for you and caches them and return the URL to a local cached texture or if it gives you the original URL
  • [13:25] FWord Utorid: and afaic the only reason UUIDs are relevant to the viewer is NONE
  • [13:25] Whump Linden: So if a texture is at a URL that isn't in an asset store, such as a flickr image, does the simulator handle that or does each viewer go and request it?
  • [13:25] Zha Ewry: That should never happen, if the URIs are well formed
  • [13:26] Teravus Ousley: yep.. one would argue that you can do that on the web also
  • [13:26] FWord Utorid: whump, the viewer requests it
  • [13:26] Tammy Nowotny: wonders if this would cause security isuses, since assets could be used to track IPs, miuch as images are used today on web pages.
  • [13:26] Zha Ewry: Depends on the model. The protcol shouldn't cae.
  • [13:26] FWord Utorid: you don't want the simulator opened up to request ... questionable material
  • [13:26] Meadhbh Oh: +1 Tammy
  • [13:26] Teravus Ousley: puts spacer.gif in the root folder.. and /images/spacer.gif in another folder.. and references both.
  • [13:26] Tao Takashi: even if it's available under 10 URIs but you only know one it wouldn't be a problem ;-)
  • [13:26] Rex Cronon: the sim gets it once, and viewer request?
  • [13:26] Tao Takashi: but the same would be true for UUIDs anyway
  • [13:26] FWord Utorid: tammy, assume everyone knows your IP
  • [13:26] Whump Linden: Yes, Tammy, I was thinking of texture as web bug.
  • [13:27] Tammy Nowotny: LOL, OK FWord, I will
  • [13:27] Zha Ewry: sighs at FWord, who is mostly right about that IP address being public
  • [13:27] FWord Utorid: in IRC, your IP is public.
  • [13:27] Tao Takashi: we have lots of VW bugs these days already.. think URLs which are requested to pop up.. moreover region operators also have that data
  • [13:27] Whump Linden: note that we have the same issue today with parcel media
  • [13:27] FWord Utorid: so I can go and peek at Zha's IP address right now
  • [13:27] Meadhbh Oh: @Tao... yeah... i guess if you wanted to improve the caching, you could hash each object (textures included) and link URLs with the same hash
  • [13:27] Meadhbh Oh: but
  • [13:27] Meadhbh Oh: that's probably premature optimization
  • [13:27] FWord Utorid: and yeah, anything which grabs from the web, exposes IP
  • [13:28] Saijanai Kuhn:  : first nakid UUID now nakid IP...
  • [13:28] FWord Utorid: the thing is this, does the second life viewer HAVE TO identify itself with it's user agent
  • [13:28] Meadhbh Oh: i wonder if this should be an agent or region setting?
  • [13:28] Whump Linden: and pushing that to the sim means you have to trust the sim, but you're rezzing there, so you have to trust the region
  • [13:28] Tao Takashi: I think the way it will (and IMHO should) go is that VWs and Web will more and more integrate with each other but that also means that more and more stuff will have your IP
  • [13:28] Meadhbh Oh: the ability to grab arbitrary content from the web
  • [13:28] Tao Takashi: but you should have controls of course not to load external textures etc.
  • [13:28] FWord Utorid: yeah, that's possible
  • [13:28] Meadhbh Oh: in one model you have the sim grabbing everything and giving it to the viewer
  • [13:28] FWord Utorid: no texture phishing
  • [13:28] Teravus Ousley: well the only issue with it being a region setting is.. format specifications.
  • [13:29] Latif Khalifa: or have a sim act as a proxy like it is done now
  • [13:29] Meadhbh Oh: in the other, the viewer is much more active... and grbs arbitrary content for itself
  • [13:29] Teravus Ousley: .. JPEG2000.. must be encoded.. and what not
  • [13:29] Tammy Nowotny: I am at [2] :-)
  • [13:29] FWord Utorid: meadhbh, there are several issues with the sim as a proxy
  • [13:29] Zha Ewry: Proxy, is in some ways much more likely to keep cinstsnecy
  • [13:29] Tao Takashi: the things is: you might want to write some application which changes the texture on the web server it is fetching it from
  • [13:29] Tao Takashi: then a cache might be unhandy..
  • [13:29] Tao Takashi: but then again you don
  • [13:29] Meadhbh Oh: funny.. i'm at [3]
  • [13:29] Zha Ewry: One problem with fetching the textures alll at the client,is..
  • [13:29] Dahlia Trimble: lol
  • [13:29] Tao Takashi: 't know when the viewer is refreshing it
  • [13:29] FWord Utorid: i host my sims on a server
  • [13:29] Zha Ewry: that you expose peple to lots of variable pathing
  • [13:29] Teravus Ousley: I'm at [4] :D
  • [13:29] Morgaine Dinova: Meadhbh: you've hijacked my IP!
  • [13:29] Tao Takashi: so probably you will change the texture URL with some script call instead
  • [13:30] Dahlia Trimble: I/m at [5]
  • [13:30] FWord Utorid: if people could make prims there that run my bandwidth costs up, i might have a huge bill
  • [13:30] FWord Utorid: I am at 10.10.10.10
  • [13:30] Meadhbh Oh: but there are several people who probably want to allow their accounts the ability to use arbitrary content
  • [13:30] Strawberry Fride:  ::1, I think
  • [13:30] Whump Linden: and I'm at the end of a bunch of Her Majesty's Signal Corps doing IP over sempahore.
  • [13:30] Meadhbh Oh: waxes poetic for teh days when 10.x was IBM
  • [13:30] Tao Takashi: 11. was Tao :)
  • [13:30] FWord Utorid: acknowledges this as the first time we've seen Tess's actual head.
  • [13:30] Meadhbh Oh: wow... that's some astonishingly good throughput for semaphors...
  • [13:30] FWord Utorid: [everyone stares at tess's head]
  • [13:31] Whump Linden: okay, thinking about assets like textures on the web
  • [13:31] Latif Khalifa: never underestimate the bandwith of a truck full of hard disks :P
  • [13:31] Morgaine Dinova: Sneakernet
  • [13:31] Zha Ewry: 747s full of tapes
  • [13:31] FWord Utorid: wait, nvm, that was the wrong person, her head is still obscured. ctrl-shift-alt 1
  • [13:31] Whump Linden: what about services that domains provide that are going to be consumed as traditional web services?
  • [13:32] Meadhbh Oh: a ballot box full of chad
  • [13:32] Whump Linden: that is, not by viewers, but browsers and mashups?
  • [13:32] Morgaine Dinova: Latency of sneakernet sucks for realtime audio
  • [13:32] Tao Takashi: I wish SL today would have such web services ;-)
  • [13:32] Zha Ewry: Ah. Now you're talking QOS, Morgaine.
  • [13:32] Bartholomew Kleiber: depends on the guitar
  • [13:32] Tao Takashi: more of them that is
  • [13:32] Meadhbh Oh: you might get different semantics depending on whether you want the sim to be a proxy for ALL the agents in the sim
  • [13:32] FWord Utorid: when we can run RDP on a prim, I'll be happy
  • [13:32] Latha Serevi: I wish I understood the (seemingly very basic) issue of when a URL should point to a constant entity (and you must change the URL to make even a small change to the entity) and when a URL should signify a constant "role" (and its contents are changeable). Can anyone point me to a useful discussion?
  • [13:32] Whump Linden: Morgaine, what about a taper at a dead show?
  • [13:32] Tao Takashi: I guess you can see world.secondlife.com as one
  • [13:33] Latif Khalifa: Whump, don't you think the priority for OGP would be to get the viewer working and worry about other stuff later?
  • [13:33] Whump Linden: yes, Tao
  • [13:33] Meadhbh Oh: or you want the individual viewers to get their own connection to the resource.. .using the sim simply to connect the viewer to the resource
  • [13:33] FWord Utorid: is there actually a lot of viewer mod going on to support ogp?
  • [13:33] Tao Takashi: I also wonder if the situation with scripts will stay as it is. Maybe many of those bigger applications will be more replaced by server plugins
  • [13:33] lyndell Aleixandre: in the <sim caches assets from the web and acts like a proxy> context .. its ineficient use of bandwidth and open to abuse, as for the other solution viewer fetching assets by itself off the web .. opens security concerns..
  • [13:33] Tao Takashi: I mean most of the time LSL scripts are more a hack than anything else
  • [13:33] Zha Ewry: REST implies that if the state changes, its a new URL
  • [13:33] Meadhbh Oh: +1 server plugin concept
  • [13:33] Morgaine Dinova: Whump: the dead would be the proto Creative Commons ;-)
  • [13:34] Tao Takashi: but if you can run your own region then it would be much easier to run your own stuff there
  • [13:34] FWord Utorid: there's already server plugin potential with OpenSIm
  • [13:34] Tao Takashi: like web apps on your own web server
  • [13:34] Teravus Ousley: +1 on Server plugin concept
  • [13:34] Zha Ewry: one way to mark he distinctoin, is to seperate REST resources from the non REST static stuff
  • [13:34] Whump Linden: but a static resource is still a resource
  • [13:34] Tao Takashi: one you can only GET I assume
  • [13:35] Teravus Ousley: notes that some of that time will be eaten up by encoding JPEG2000 files though
  • [13:35] Tao Takashi: but then.. you might also want to upload such a resource.. now it depends on the definition of static ;-)
  • [13:35] Teravus Ousley: just as going the other way.. there's also some decoding time..
  • [13:35] lyndell Aleixandre: unless the viewer or sim will only decode or fetch jpeg2000 encoded images
  • [13:35] FWord Utorid: i am prone to wonder why jp2k is required
  • [13:35] Meadhbh Oh: j2k is legacy
  • [13:36] lyndell Aleixandre: because its in the viewer
  • [13:36] Teravus Ousley: I'm only aware of the viewer using jpeg2000
  • [13:36] Tao Takashi: I think in SL it's mostly used because it's streamable
  • [13:36] FWord Utorid: if this is going to be a portal to accessing different media types then that will need to be examined
  • [13:36] Teravus Ousley: yep.. that would be my guess also Tao
  • [13:36] Whump Linden: FWord, I think it is a legacy thing, but have to ask. Tess?
  • [13:36] Meadhbh Oh: in theory.. we could redefine the standard as png, tiff, eps or whatever
  • [13:36] FWord Utorid: there shouldn't be a 'standard', it should be able to view all media types
  • [13:36] Meadhbh Oh: but i'm going to add svg before we do that
  • [13:36] Tao Takashi: but then again there is probably nothing to prevent PNGs from being used, it just might take a bit more time until you see the texture
  • [13:36] FWord Utorid: even power point and pdf
  • [13:36] Bartholomew Kleiber: compuserve GIFs *ducks
  • [13:37] Meadhbh Oh: tao.. right
  • [13:37] Teravus Ousley: well you could even throw them in the current UDP stream.. but you'd have to make use of more image format uints
  • [13:37] Tao Takashi: but even with allowing any type not every viewer will support every type out there
  • [13:37] Morgaine Dinova: Which brings up the question: in OGP, are images going to have a mime type in OGP? If not then we have to standardize on jpeg2k or whatever, which would be pretty bad for evolution.
  • [13:37] Teravus Ousley: .. there's a space in the packet for 'type'
  • [13:37] FWord Utorid: teravus, i am just thinking of long term desktop applications and such
  • [13:37] Teravus Ousley: .. but I've only seen jpeg2000 used
  • [13:37] FWord Utorid: leveraging SL as a higher level interface
  • [13:37] Tao Takashi: I would assume that images are downloaded via HTTP and thus can have a content-type
  • [13:37] Meadhbh Oh: wait wait... i thought png supported progressive download
  • [13:38] Tao Takashi: right, I am no PNG expert ;-)
  • [13:38] Tao Takashi: so even better
  • [13:38] lyndell Aleixandre: seems to me your going to be using the <viewer fetches asset data from URL> approach instaid of the proxy approch
  • [13:38] FWord Utorid: anyway, that's extremely far future thinking, i don't want to derail the important discussion at hand with format woes
  • [13:38] Meadhbh Oh: @Morgaine... one would hope that the media type of an image you get from teh simulator would identify it's format
  • [13:38] Tao Takashi: viewer fetches texture from sim cache (also a URL) which fetches it from whereever
  • [13:38] Teravus Ousley: yep, there's a uint in the packet for image type
  • [13:39] Whump Linden: PNG supports interlacing.
  • [13:39] Meadhbh Oh: eek! i just saw the duke on Rex's shoulder
  • [13:39] Tao Takashi: it again would be sort of implementation specific if the viewer caches it or not
  • [13:39] FWord Utorid: the java guy
  • [13:39] Tao Takashi: viewer just gets a URL and downloads it, get's a content-type header with it
  • [13:39] Meadhbh Oh: @Terevus... but in the brave new OGP future... everything that can be http will be http (or so i thought)
  • [13:39] FWord Utorid: go http
  • [13:39] Teravus Ousley: preferrably :D
  • [13:39] FWord Utorid: that's totally win
  • [13:40] lyndell Aleixandre: *sigh* protocol overheads...
  • [13:40] Tao Takashi: it will all be rhttp ;-)
  • [13:40] Teravus Ousley: Images should definately be HTTP
  • [13:40] lyndell Aleixandre: i agree .. as for everything else ...
  • [13:40] Meadhbh Oh: wonders why there's not a Java simulator.. then gets back to thinking about pressing issues
  • [13:40] FWord Utorid: *sigh* computers will get faster and it won't matter
  • [13:40] Tao Takashi: installs mono
  • [13:40] lyndell Aleixandre: yes fw computer speed solves all code inefficencies
  • [13:40] Rex Cronon: who can translate the viewer to java:)
  • [13:40] Meadhbh Oh: well... has anyone done the math on how much overhead you get with HTTP vs. images ov UDP?
  • [13:41] Whump Linden: lyndell: so do we have a custom transport, some messages using UDP?
  • [13:41] FWord Utorid: the overall flexibility of maneuvering to an http base and the opportunities it opens up outweigh the processing issues
  • [13:41] lyndell Aleixandre: not nessesarily
  • [13:41] Tao Takashi: I want one in Python ;-)
  • [13:41] Tao Takashi: http also means more developerts
  • [13:41] Meadhbh Oh: cause i gotta say... the "let's just recreate a bad vesion of TCP using UDP" in the existing viewer is.. uh...
  • [13:41] Tao Takashi: -t
  • [13:41] Meadhbh Oh: suboptimal
  • [13:41] FWord Utorid: inherently, only controls really require udpish stuffs
  • [13:41] Meadhbh Oh: using TCP is probably "a good thing"
  • [13:41] Teravus Ousley: probably 1.5 to 2 times the bandwidth.. but.. it only gets sent once.. instead of 20 times :D
  • [13:41] Rex Cronon: cosidering that right now is "so easy" to ready the existing code:)
  • [13:41] lyndell Aleixandre: whump .. would you use http for the viewer to communicate object interpolation deta?
  • [13:42] Meadhbh Oh: and if you're gonna use TCP, adding HTTP doesn't really add too much overhead ( i would think )
  • [13:42] Teravus Ousley: (refering to the UDP Resend)
  • [13:42] Latif Khalifa: you still need UDP
  • [13:42] Morgaine Dinova: Zha: are textures grabbed by the viewer using ordinary webby image access in our brave new world, or are such objects streamed like other client objects?
  • [13:42] FWord Utorid: udp is only needed for controls, if that
  • [13:42] Tao Takashi: and we also add caps roundtrips to that actually
  • [13:42] Dahlia Trimble: I kinda like the idea to use UDP for object motion
  • [13:42] Dahlia Trimble: not much else thouth
  • [13:42] FWord Utorid: yeah, and updates, as dahlia said
  • [13:42] Latif Khalifa: you don't want to have a flying physical object passing around you and getting the object updates over TCP trust me
  • [13:43] Meadhbh Oh: well.. the "reliable" UDP in the source base now definitely increases teh complexity of the code
  • [13:43] Meadhbh Oh: @Latif... why not?
  • [13:43] lyndell Aleixandre: does anyone know why we cant use http for object updates
  • [13:43] Tao Takashi: yep, UDP adds quite a bit to the complexity of the messaging system in pyogp also
  • [13:43] Meadhbh Oh: if you mention latency and lost packets
  • [13:43] Tao Takashi: and I am a bit afraid of touching it
  • [13:43] Dahlia Trimble: lose the reliabile flag and send reliable packets over TCP
  • [13:43] Whump Linden: lyndell, I'm not conversant with how object interpolation deltas currently works, so don't have a answer for that.
  • [13:43] Meadhbh Oh: i'll bring up the fact that probably no one here has tried using a different congestion algorithm
  • [13:43] FWord Utorid: object updates are the time sensitive component, everything else has a little bit of room
  • [13:43] Rex Cronon: i think it was thought that UDP is faster
  • [13:44] Tao Takashi: what if object updates are streamed via TCP?
  • [13:44] Meadhbh Oh: in my previous job, we did rtp over TCP
  • [13:44] Meadhbh Oh: and it was
  • [13:44] Meadhbh Oh: well... a touch weird
  • [13:44] Meadhbh Oh: and there's definitely overhead for the three-way handshake
  • [13:44] Latif Khalifa: such rapid object updates need not be sent over reliable links, and TCP's error checking and packet reordering adds a lot of unneccssary overhead
  • [13:44] Meadhbh Oh: but we played around with congestion and found somethign taht actually worked
  • [13:44] lyndell Aleixandre: my concern is .. http adds conciderable latency, and although it works out great for some tasks .. it does not solve all
  • [13:44] Morgaine Dinova: Forget TCP for messaging, there's a better kid on the block with all the benefits of UDP but none of the disadvantages: SCTP
  • [13:45] Meadhbh Oh: @Latif... do we have data on how much error checking and packet reordering adds?
  • [13:45] Tao Takashi: well, I agree that having not reliable packets might also make the code lighter
  • [13:45] Bartholomew Kleiber: Meadhbh: which was?
  • [13:45] Whump Linden: I'm trying to remember the URL for it, but Live Journal publishes a continious spew of public entries URL somewhere, and I have to remember how you consume it.
  • [13:45] Teravus Ousley: well, with TCP you have flow control... UDP.. you do not :D
  • [13:45] Whump Linden: looks at google.
  • [13:45] Teravus Ousley: better control of the Quality of service
  • [13:45] Meadhbh Oh: okay... how much latency does TCP add?
  • [13:45] Latif Khalifa: and you don't need or want flow control for certain types of messages
  • [13:45] Meadhbh Oh: if you eliminate slow-start
  • [13:46] Latif Khalifa: Meadhbh, a lot, I know of no online game (using game server or some such) that does not use UDP
  • [13:46] Meadhbh Oh: i think what i'm getting at is... i don't think we have a lot of data
  • [13:46] lyndell Aleixandre: QOS is something that should be looked at ... such things as throttling etc...
  • [13:46] Teravus Ousley: subtract images.. and you're right..
  • [13:46] FWord Utorid: ultimately, one should scale a service to facilitate 10 X what one wants, rather than try to make it run in 10% of the time
  • [13:46] Whump Linden: not finding it offhand, will look for it later
  • [13:46] Tao Takashi: the question is if we can make it to work with TCP but it's not HTTP and the TCP code is as complex as the UDP one right now if that brings much ;-)
  • [13:46] Latif Khalifa: Meadhbh, there is plenty of data about it... its the reason UDP still exist
  • [13:46] Meadhbh Oh: post hoc ergo propter hoc
  • [13:46] Morgaine Dinova: [6]
  • [13:46] Teravus Ousley: images consists of 85-90% of the traffic
  • [13:47] Meadhbh Oh: aggghhhh! someone mentioned SCTP
  • [13:47] FWord Utorid: yeah, the prim definitions and the updates are very minimal
  • [13:47] Meadhbh Oh: actually... i would love to see sctp more widely supported
  • [13:47] FWord Utorid: although, one definately wants updates to be as expedient as possible, the rest is images
  • [13:47] FWord Utorid: and images, though you want them quick, don't have to come over the update pipe
  • [13:47] Tao Takashi: just forbid flying objects ;-)
  • [13:47] Latif Khalifa: all I'm saying is that for stuff like object updates you *want* to use udp ;)
  • [13:47] Morgaine Dinova: It's fully supported --- it's just not used a lot, because people are hung up on UDP/TCP
  • [13:47] FWord Utorid: latif, i agree
  • [13:47] Tao Takashi: it might be very welcomed on certain interviews as well ;-)
  • [13:48] Teravus Ousley's: feelings on the matter are fully documented on ugotrade now :D
  • [13:48] Meadhbh Oh: @Morgaine... there are a lot of router manufacturers that don't support SCTP fully (unfortunately)
  • [13:48] Meadhbh Oh: but i wonder if that's still the case
  • [13:48] Teravus Ousley: [7]
  • [13:48] Meadhbh Oh: @Latif... why?
  • [13:49] FWord Utorid: in terms of the architecture, i think the real latency issues are caused mostly by sim crossing, something that could be maneuvered by having user services simulators rather than area simulators
  • [13:49] Latif Khalifa: Meadhbh, latency
  • [13:49] Morgaine Dinova: Meadhbh: backbone manufacturers like Cisco all support it. But as you say, some personal routers might not
  • [13:49] lyndell Aleixandre: cool .. so .. http for textures/caps/primative assets .. sctp for updates/physics/IM?
  • [13:49] Meadhbh Oh: there's no data that says that object updates will get to your viewer any faster by using UDP
  • [13:49] FWord Utorid: why sctp?
  • [13:49] Latif Khalifa: Meadhbh, but there is
  • [13:49] Tao Takashi: sctp will decrease the developer acceptance I'd say
  • [13:49] FWord Utorid: because it's a new acronym we've never heard of!
  • [13:50] Meadhbh Oh: @Morgaine... no... some of the product lines from Cisco & Juniper sometimes produce patches that break things like throttling over SCTP
  • [13:50] Rex Cronon: meadhbh, u could be the one that proves that:)
  • [13:50] Meadhbh Oh: what i mean is taht you often get featurs for UDP and TCP well before SCTP
  • [13:50] Morgaine Dinova: Because it supports messaging directly but without any of the problems of UDP, and it has *VERY* small latency owing to multiple parallale bundles.
  • [13:50] Meadhbh Oh: @Rex... ugh... i might have to be
  • [13:50] Dahlia Trimble: the problem with sending object updates reliably is a lot of effort can go into guaranteeing delivery of packets that are no longer meaningful
  • [13:51] Rex Cronon: how can u have parallel bundels when there is only one wire?
  • [13:51] Tao Takashi: found some pysctp from 2006
  • [13:51] Meadhbh Oh: @Dahlia.. thi is exactly the same problem you get with RTP
  • [13:51] Tao Takashi: alpha version
  • [13:51] Tao Takashi: let's take it :)
  • [13:51] Dahlia Trimble: which makes UDP attractive for that scenario
  • [13:51] Meadhbh Oh: and... as i said... in some cases... you can do RTP over TCP with no loss of QoS
  • [13:51] FWord Utorid: UDP is sort of 'proven' for SL
  • [13:52] Morgaine Dinova: Rex: in the same way as we transmit bytes when there is only one wire :-) What makes the operation of bundles concurrent is that delay in one bundle doesn't affect another.
  • [13:52] Meadhbh Oh: umm... UDP is sorta proven to obfuscate the codebase
  • [13:52] Teravus Ousley: There is the 'proven factor'
  • [13:52] Teravus Ousley: heh
  • [13:52] FWord Utorid: so it just comes down to what you 'drop' from that to maneuver to using web services and maintain the status quo
  • [13:52] Bartholomew Kleiber: but thats not UDPs fault
  • [13:52] Meadhbh Oh: but yeah... this is the type o thign that really needs a touch of research
  • [13:52] Dahlia Trimble: lol
  • [13:52] Meadhbh Oh: +1 Bart
  • [13:52] Meadhbh Oh: yup
  • [13:52] Latif Khalifa: Meadhbh, yeah do the research, you will find plenty on the web
  • [13:53] Latif Khalifa: people don't implement UDP based messaging just for the fun of it
  • [13:53] Meadhbh Oh: Latif... i did the resarch for RTP
  • [13:53] Meadhbh Oh: and as i said
  • [13:53] Meadhbh Oh: killing slow-start on your connections
  • [13:53] Latif Khalifa: it is the best solution for certain class of problems out there
  • [13:53] FWord Utorid: ultimately, the 'ultra-tight protocol that runs on a 1 mhz processor' is never what wins, but what wins is 'what will developers code for'
  • [13:53] Meadhbh Oh: gives you performance that is just as good as udp
  • [13:53] Dahlia Trimble: or what will sell more high end hardware
  • [13:53] Meadhbh Oh: as long as you're below the throughput rate
  • [13:53] FWord Utorid: dahlia, precisely
  • [13:53] Latif Khalifa: and doing anythng "exotic" that you $20 netgeaer router does not do is never going to fly either
  • [13:54] Meadhbh Oh: above the throughput rate.. .both UTP and TCP have problems
  • [13:54] Latif Khalifa: so you need to stick to UDP and TCP
  • [13:54] Teravus Ousley: thinks developers will code for lots of things assuming there's incentive and a library that they can plug into their application
  • [13:54] Meadhbh Oh: +1 FWord
  • [13:54] FWord Utorid: the only items i haven't heard mentioned that could be beefy are animation, voice, and video
  • [13:54] Meadhbh Oh: personally.. i like the idea of simplifiying thigns 'til it's easy to figure out where the problems are
  • [13:54] lyndell Aleixandre: FW i have a 3.2 ghz athlon 64 processor ... and you know what i notice? the framerate starts dropping after my bandwidth hits 500 kbps ... why is that?
  • [13:54] Meadhbh Oh: and _then_ optimizing
  • [13:54] FWord Utorid: which, really, sl's protocols will just be the pointer towards services using those
  • [13:55] FWord Utorid: lyndell, you are a resident of a country which was made as a penal colony
  • [13:55] Morgaine Dinova: Fixing the network protocol is actually the wrong approach altogether. We should be working at a layer above that, or else our infrastructure will be stuck in the past.
  • [13:55] Meadhbh Oh: +1 Morgaine
  • [13:55] Meadhbh Oh: but
  • [13:55] Meadhbh Oh: iwth the caveat
  • [13:55] Dahlia Trimble's: bandwidth cap is set at 250
  • [13:56] Meadhbh Oh: taht you probably want to allow the "layer above" to make recommendations to the "layer below"
  • [13:56] Whump Linden: Ugh, Comcast, that's right.
  • [13:56] Meadhbh Oh: Whump! 5L$ in the swear jar! C**cast is a curse word in some parts!
  • [13:56] Whump Linden: And unlike Telstra, they don't do a deal for SL.
  • [13:56] lyndell Aleixandre: so now we are building a protocol stack
  • [13:56] Tammy Nowotny: I forgot to pay my DSL bill today
  • [13:56] Morgaine Dinova: Meadhbh: sure, via socket options or whatever. That can be done.
  • [13:57] Meadhbh Oh: @lyndell.. uh... we've already made a protocol stack
  • [13:57] lyndell Aleixandre: lol i know...
  • [13:57] Meadhbh Oh: what i'm saying is why not use the protocol stack that was developed for NET4
  • [13:57] FWord Utorid: ultimately, this needs to work the way that will make it possible to do more useless crap that looks cool and i can brag to my friends about. efficiency of it doesn't matter, it just has to be awesome.
  • [13:57] Tao Takashi: maybe I should be glad to have Telekom DSL ;-)
  • [13:58] Meadhbh Oh: and understood by legions of grad students
  • [13:58] Dahlia Trimble: whatever is used shoud be subjected to a *lot* of testing
  • [13:58] Meadhbh Oh: (and presumably understood by ISPs)
  • [13:58] Tao Takashi: yay, opensim build failed.. :-/
  • [13:58] Whump Linden: Meadhbh, NET4 is?
  • [13:58] FWord Utorid: if it doesn't look cool, i will switch to another service that looks cool.
  • [13:58] Meadhbh Oh: the implementation that was in the BSD clones of the distant past
  • [13:58] Teravus Ousley: yep.. should be understood by ISPs hopefully.. so they won't re-prioritize the packets :P
  • [13:58] lyndell Aleixandre: lol FW .. MUD is for you
  • [13:59] Meadhbh Oh: uh oh... peeps are gonna grief me for calling Linux a BSD clone
  • [13:59] lyndell Aleixandre: blasphamy!
  • [13:59] FWord Utorid: by looking cool, i mean, this should be where i can rez a computer and connect that computer to more computers
  • [13:59] Meadhbh Oh: @Terevus.. well... all we have to do is simply move over to v6 and then we get QoS for free, right?
  • [14:00] FWord Utorid: i should be able to watch a music video in whump's little gearview window, backwards
  • [14:00] Morgaine Dinova: Everything's a clone of Unix V7 from 1979, get over it ;-)
  • [14:00] Tara5 Oh: Gotta to run cyas all soon
  • [14:00] Whump Linden: okay, it's 2pm, thanks everyone for coming over.
  • [14:00] lyndell Aleixandre: awesomeness
  • [14:00] Meadhbh Oh: nah... everything's a clone of TENEX
  • [14:00] Whump Linden: Zero should be back on Thursday at 8:30 AM SLT.
  • [14:00] Rex Cronon: bye whump
  • [14:00] Morgaine Dinova: Thanks Whump
  • [14:00] Rex Cronon: bye tara
  • [14:00] Strawberry Fride: thanks Whump :)
  • [14:00] Meadhbh Oh: okay... so... takeaways?
  • [14:00] FWord Utorid: thanks for the free stuff
  • [14:00] Meadhbh Oh: i should go run some numbers re: UDP and TCP.
  • [14:01] Meadhbh Oh: i actually suspect there will be some places where UDP "will be the answer"
  • [14:01] Latif Khalifa: Meadhbh why do you think every game designer uses UDP`
  • [14:01] Tammy Nowotny: TY Whump... very interesting :-)
  • [14:01] Thoys Pan: take a seat brother in law
  • [14:01] Tao Takashi: will keep working on making opensim build
  • [14:01] Meadhbh Oh: but I also think we're going to find taht TCP is surprisingly good on some tasks we thought it might not ne
  • [14:01] Meadhbh Oh: be
  • [14:01] Rex Cronon: bye everybody
  • [14:01] Whump Linden: and some questions about PNG's streamability
  • [14:01] Morgaine Dinova: Cya Rex
  • [14:01] Teravus Ousley: take care
  • [14:01] Bartholomew Kleiber: bye
  • [14:01] Dahlia Trimble: bye all :)
  • [14:01] Meadhbh Oh: @Latif... cause game developers are in some ways prisoners of their experience
  • [14:02] FWord Utorid: bye dahlia
  • [14:02] Latif Khalifa: Meadhbh, wrong
  • [14:02] Meadhbh Oh: i've yet to see a game developer do network research
  • [14:02] Latif Khalifa: Meadhbh, because its the best solution for certain class of problems
  • [14:02] Meadhbh Oh: and you have the numbers to verify this?
  • [14:02] Tao Takashi: does opensim work with mono2.0?
  • [14:02] FWord Utorid: i can invent numbers
  • [14:02] Latif Khalifa: Meadhbh, there is a ton of it outhere
  • [14:02] Latif Khalifa: Meadhbh, just google it :P
  • [14:02] Meadhbh Oh: like?
  • [14:02] Morgaine Dinova: Latif: our needs are beyond the capabilities of UDP, that's the problem. Like Zero said, there's a large semantic gap.
  • [14:03] lyndell Aleixandre: some of the issues raised today are speed vs programability , and although http might be easy to use allot of people are transparent to the fact of latency problems and protocol overheads that can negatively affect the preformance of both the viewer and simulator .. many people dont care about this semantic though.
  • [14:03] FWord Utorid: that's nice, lyndell, but you're arguing with the people who are going to make it
  • [14:03] FWord Utorid: and they are going to make it the http way
  • [14:03] lyndell Aleixandre: thats cool, im not trying to convince anyone otherwise
  • [14:04] Meadhbh Oh: all i'm saying is we should revisit the discussion of UDP vs. TCP.
  • [14:04] lyndell Aleixandre: stuff like textures are great for http
  • [14:04] Latif Khalifa: Morgaine, I am not saying UDP is sufficient for everything. I am saying you don't want to make everything run over TCP or god forbit some protocol implemented on top of HTTP
  • [14:04] Latif Khalifa: it would make real time simulation useless
  • [14:04] Meadhbh Oh: there are currently several VoIP products that do RTP over TLS over TCP
  • [14:04] Whump Linden: Saij, do you do the transcripts for Zero, or I should I send him a notecard?
  • [14:04] Meadhbh Oh: and they perform considerably better than their UDP equivalents
  • [14:04] Saijanai Kuhn:  : I'll post
  • [14:04] Whump Linden: kk, thanks Saij.
  • [14:04] Meadhbh Oh: whether the existing SL protocol would benefit from being over TCP? i dunno.
  • [14:05] Meadhbh Oh: but. we shold do the reesarch
  • [14:05] Saijanai Kuhn:  : still keeping it rolling btw
  • [14:05] Latif Khalifa: Meadhbh, do that
  • [14:05] Zha Ewry: Well, over http, not udp,
  • [14:05] Whump Linden: okay, I need to run. Thanks again, everyone.
  • [14:05] Morgaine Dinova: Latif: well that's why SCTP was developed, to provide UDP-type low latency with TCP-type reliability and without the straightjacket of UDP for messaging.
  • [14:05] Latif Khalifa: see you Whump
  • [14:05] Meadhbh Oh: yup... i'm running too.
  • [14:06] Whump Linden: Meadhbh, if you research this, putting the results on the wiki would be good.
  • [14:06] Latif Khalifa: Morgaine, does it run on $20 D-Link?
  • [14:06] Morgaine Dinova: Latif: ask D-Link
  • [14:06] Meadhbh Oh: yup.
  • [14:06] Zha Ewry: all 500 variations of ht $30 d-link, of course
  • [14:06] FWord Utorid: it's a hard sell, SCTP
  • [14:06] Zha Ewry: 'tis, with reason
  • [14:07] Latif Khalifa: I don't see how to allow SCTP through my router
  • [14:07] FWord Utorid: whether it's better or not, it's like going to Microsoft and telling them to run SUN
  • [14:07] Latif Khalifa: so i guess its not supported
  • [14:07] Zha Ewry: Or it may work fine, and you just don't know
  • [14:08] FWord Utorid: well, i've stared at a computer for a long time, and never heard of it, so SCTP has crappy marketing
  • [14:08] Latif Khalifa: and low support in commodity hardware to go with it i guess
  • [14:08] Morgaine Dinova: Half the personal routers in the world are going to have to be replaced for IPv6 anyway, and we can't ignore that coming. So it's all change in nettworking folks.
  • [14:08] Tao Takashi: there also seems to be no real support for at least python for SCTP
  • [14:08] Latif Khalifa: Morgaine, that's 10 years from now
  • [14:08] Tao Takashi: not sure how it looks with different languages
  • [14:08] Zha Ewry: Sure, but you want to follow that change, not lead, it Morgaine
  • [14:09] FWord Utorid: never heard of .net SCTP, but like i said, just heard of SCTP
  • [14:09] Tao Takashi: me, too ;)
  • [14:09] Tao Takashi: well, I might have heard of it but have forgotten about it again
  • [14:09] Morgaine Dinova: Nope Latif. You may still be using IPv4, but since addresses will run out, there will be people on IPv6, so VWs have to support it.
  • [14:09] Tao Takashi: and I would prefer to use something for networking which is in the standard lib
  • [14:09] Tao Takashi: which means well tested
  • [14:09] Latif Khalifa: so its not a realistic option for a protocol that is supposed to be use by "ordinary people"
  • [14:10] Morgaine Dinova: SCTP support is on all platforms, and all the main languages reputedly have bindings, although I haven't checked.
  • [14:10] Tao Takashi: gna, now my mono installation is completely borked I guess
  • [14:10] Morgaine Dinova: Hehe
  • [14:10] Zha Ewry: ok. time to run to the 2:00 and be 10 minute slate