User:Zero Linden/Office Hours/2007 Jul 24

From Second Life Wiki
Jump to: navigation, search

Transcript of Zero Linden's office hours:

[13:05] Zero Linden: hello hello
[13:05] Kooky Jetaime: Hi Zero
[13:05] Zha Ewry: Afternoon Zero
[13:05] Zero Linden: sorry all for being a little late
[13:05] Adam Xinpeng: Hello'
[13:05] otakup0pe Neumann: g'day
[13:05] Kooky Jetaime: We got tons of techy questions for ya today..
[13:05] McCabe Maxsted: ahoy zero
[13:05] Feep Larsson: Hey Zero
[13:05] McCabe Maxsted: how was lunch?
[13:05] Kooky Jetaime: maybe
[13:06] Zero Linden: lunch was lovely - Indian dal and chana
[13:06] otakup0pe Neumann: Sounds delightful.
[13:06] Kooky Jetaime: The first being the best one, courtesy of McCabe, which is who should we ask about a missing text chat link?
[13:06] otakup0pe Neumann: mmm curry. brb.
[13:07] Zero Linden: oooo - a new sheep! Hello Feep
[13:07] otakup0pe Neumann: Feep the Sheep
[13:07] otakup0pe Neumann: Heh. Was that intentional ? ;)
[13:08] Adam Xinpeng: Oh, <--- Adam Zaius's secret alt (I'm logged in there too, waiting on .RAW's to upload)
[13:08] Zero Linden: well - first off, welcome all to my geeky office hours.
[13:08] McCabe Maxsted: the support site says premium accounts have access to live text chat, but there's no live text chat to be found
[13:08] Feep Larsson: Nope, just happenstance! Hello Zero!
[13:08] Zero Linden: As always, we'll be discussing the technical undepinnings (underthings?) of SL
[13:08] Kooky Jetaime: And my question is a lot more geeier :)
[13:08] Kooky Jetaime: geekier
[13:08] Zero Linden: And as always, it'll all make it to the public wiki
[13:09] Zero Linden: missing "text chat link" - I'm sorry.... Context... anyone have any context here? :-)
[13:09] otakup0pe Neumann: Hehe. Support portal.
[13:09] Kooky Jetaime: Support Portal
[13:09] Kooky Jetaime: The web page prior to portal login says premium users have access to text chat
[13:09] Wyn Galbraith holds still to see if she gets to stay this time.
[13:09] Kooky Jetaime: however inside the portal, there is only the link for Concerge customers
[13:09] Kooky Jetaime: which us premies can't see
[13:10] Zero Linden: like as in access to text chat on the web page?
[13:10] Kooky Jetaime: and Since its not your speciality, who should we talk to.
[13:10] Zero Linden: what? never heard of such a thing
[13:10] Zero Linden: oh oho oh
[13:10] otakup0pe Neumann: Kooky file a support ticket.
[13:10] Zero Linden: text chat on the supoort file
[13:10] otakup0pe Neumann: Is probably the best bet ?
[13:11] Kooky Jetaime: A support ticket asking for support.. ok.. am I da only one seeing irony in that?
[13:11] otakup0pe Neumann: :s
[13:11] Adam Xinpeng: Haha
[13:11] Kooky Jetaime: suppport On support I mean
[13:11] Adam Xinpeng: Meta-support. :D
[13:11] McCabe Maxsted: should I create a jira issue, do you think?
[13:11] otakup0pe Neumann: There is even an option to file one as guest, if your account is mucked up.
[13:11] Zha Ewry sees the irony.. But.. If it works..
[13:11] Kooky Jetaime: Actually yea, jira issue might be bets
[13:11] Kooky Jetaime: best
[13:11] Kooky Jetaime waits to see what zero says
[13:11] Zero Linden: Kooky - I really don't know
[13:12] Zero Linden: about how the support works
[13:12] otakup0pe Neumann: That means flip a coin. Heads support ticket, tails Jira.
[13:12] Zero Linden: I have this goofy half recollection that after you try to find a fix via the web page tools, at the end you get the option for tickets or chat or whatever premiums get....
[13:12] Zero Linden: but I'm not sure
[13:12] Zero Linden: ticket!
[13:13] Saijanai Kuhn: Didn't things change recently?
[13:13] Kooky Jetaime: Ok.... was worth a shot, my "geek" question, and this is probably stupidly simple, but how does the asset server lose an asset? (not dissapear from inventory but the "xxx missing from database" error) and why does a missing script still execute?
[13:13] Adam Xinpeng: Yeah, I think that makes sense, you need to get all the way to filing an issue before you get the live support prompt.
[13:13] Zero Linden: okay - but, now the geeky question - give me something I *can* answer
[13:13] otakup0pe Neumann: That's a good one heh.
[13:13] Adam Xinpeng: Heh, can I ask the motherlode of questions?
[13:14] Zero Linden: Well - wait, Adam, one asset at a time
[13:14] otakup0pe Neumann: haha.residents are assets too !
[13:14] Adam Xinpeng: Hehe.
[13:14] Zero Linden: So, the asset server, as you may or may no know is really a big, distributed, redudant file store
[13:14] Squirrel Wood: [8:17] Gesture is missing from database. <= would be nice if these could tell *which one* it is :)
[13:14] Zero Linden: and in that sense, it isn't going to loose an asset.....
[13:14] Adam Xinpeng: Are you still running that gargantuan garbage collector?
[13:15] Zero Linden: But yes, as AX noted, there is a Garbage Collector
[13:15] Zha Ewry observes that, while Zero is right, in a technical sense, from a user perspective, things go missing all the time
[13:15] Wyn Galbraith sees a lot of half-Ruths
[13:15] Zero Linden: for various reasons, mostly be cause we are scared, when the garbage collector (which takes weeks to complete a run) finds non-referenced assets
[13:15] otakup0pe Neumann: yeah. i get the missing gesture thing all the time on login.
[13:15] Kooky Jetaime: Redundant? not very..
[13:15] Zero Linden: they are moved to a different sub-tree on the file server
[13:16] Zero Linden: then - if they are ever accessed in the coming weeks, they are moved back
[13:16] Adam Xinpeng: I seem to remember hearing a long long time ago that occasionally when the DB's are overloaded, the garbage collector isnt able to find references to assets anymore and assumes they are unused and wipes em.
[13:16] Zha Ewry: its very redundant..
[13:16] Zero Linden: finally, after all that, they are purged
[13:16] otakup0pe Neumann: they are moved back right away Zero ?
[13:16] Zha Ewry: just not in realtime
[13:16] Zero Linden: yes - right away, actually
[13:16] otakup0pe Neumann: That might help explain why a texture or script not used for ages takes forever to show up sometimes ?
[13:16] Kooky Jetaime: Zero- this is a script that is worn in my hud every day.
[13:17] Zero Linden: but, really, it doesn't happen in practice - it is just paranoia
[13:17] otakup0pe Neumann: OK so. What does "gesture missing from inventory" mean then ;)
[13:17] otakup0pe Neumann: i totally beleive you that it's still there. but i don't think the viewer does.
[13:17] Zero Linden: if a DB is overloaded during a GC scan that involves that DB... Then we restart that scan later
[13:17] Adam Xinpeng: Means the inventory item is there, but the asset isnt.
[13:17] Zero Linden: we never let this GC run unattended
[13:17] Wyn Galbraith: It means you're running a gesture the inventory can't find.
[13:17] Wyn Galbraith: Or what Adam says ;)
[13:17] Zero Linden: we are pretty careful about actually puring an asset
[13:17] Zero Linden: No, back to the message
[13:18] Zero Linden: usually there are two possible root problems
[13:18] Wyn Galbraith has fixed that by purging her cache.
[13:18] Zero Linden: 1) There is some amount of DB fetching that is needed to deal with an asset, and if the DBs involved are borked, or overloaded
[13:18] Zero Linden: then you don't get the asset
[13:18] Zero Linden: 2) The asset server can't deliver the asset data fast enough, and then the sim gives up on the asset
[13:19] Wyn Galbraith ahs at number 2.
[13:19] Kooky Jetaime: In the case of #2, will another attempt in a few minutes most likely succeed, or does upon failure it simply give up trying?
[13:19] Zero Linden: Now, there are quite a number of different code paths for each different type of asset - and so, alas, there are a quite a number of places
[13:19] otakup0pe Neumann: So if you consistently get the message, the asset is stuck in a disused filing cabinet somewhere
[13:19] Zero Linden: I'd need to see the exact text to try and pinpoint the particualr failure mode that generated it
[13:19] Zha Ewry nods "I assume that there's not an easy way to force a retry on either, but that in time, they'll unstick?"
[13:19] Adam Xinpeng: How big is the asset cluster these days anyway? I heard back in 2006 it was around 13TB, growing at around 10%/month
[13:20] Kooky Jetaime: in a bathroom with a sign on the door sayin Beware the Leopard I presume Otakup0pe?
[13:20] Zero Linden: o.N. - more likely the DB entry that refers to it is somehow broken
[13:20] Zero Linden: like the asset field is zero'd
[13:20] Kooky Jetaime: Zero -
[13:20] Kooky Jetaime: "Script is missing from database."
[13:20] Zero Linden: AX I don't konw the current size, but I believe it is still under 20T
[13:20] otakup0pe Neumann: really ?
[13:21] otakup0pe Neumann: I was hearing over 30TB at views in dev
[13:21] otakup0pe Neumann: Either way 20TB is still a lot of pose balls.
[13:21] Adam Xinpeng: Yeah, seems actually pretty small.
[13:21] Kooky Jetaime: Ah, so if an asset has gone 0000-00... its lost forever?
[13:21] otakup0pe Neumann: So Zero, how is this dealt with ?
[13:21] otakup0pe Neumann: I'm assuming the "DB Entry" you mean is either the task or agent inventory entry ?
[13:22] Zha Ewry observes that 30TB is 60 $150 buck external USB drives. Not quite petty cash, but close.
[13:22] Zero Linden: Zha - not redundant, high-performance, distributed storage
[13:22] Zha Ewry: No, hadly
[13:22] Zha Ewry: But.. it's tiny
[13:23] Kooky Jetaime: We need some bloody redundancy...
[13:23] Zha Ewry: *hardly
[13:23] Kooky Jetaime: objects shouldn't 00000-000 for no cause
[13:23] Zha Ewry: Just pointing out, that you could stack it all on single shelf.
[13:23] Adam Xinpeng: 0000-0000 implies an inventory screwup
[13:23] Adam Xinpeng: not an asset one
[13:23] Kooky Jetaime: well
[13:23] Zha Ewry: 30TB isn't a seriusly large amount of data these days
[13:24] Zero Linden: Kooky - so every asset is stored at least twice on two different nodes, and every database is backed up by two live slaves
[13:24] Kooky Jetaime: if you have a "Script Missing from Database" and 0000-0000... is it an asset or an inventory problem? or both?
[13:24] Adam Xinpeng: "Script Missing from Database" = Asset screwup, 0000-000 = probably inventory?
[13:24] otakup0pe Neumann: ...sounds like inventory. i'm curious as to what could be done about it. i just ignore them.
[13:25] Adam Xinpeng: But, a bad inventory item could be the root cause of either.
[13:25] Adam Xinpeng: which in turn, is probably a DB problem.
[13:25] Kooky Jetaime: What about "Script Missing from Database" that is not in your inventory, but inside a rezzed object
[13:25] Adam Xinpeng: Still essentially the same thing.
[13:25] Kooky Jetaime: I have like 5 copies of my flight script, and every one of them is missing
[13:26] otakup0pe Neumann: yeah same thing kooky.
[13:26] Kooky Jetaime: Yet, the script is still executing
[13:26] Adam Xinpeng: Well that's definetely an asset issue.
[13:26] Adam Xinpeng: There's two parts to your "script"
[13:26] otakup0pe Neumann: the copies of the script are all pointing back to the same "asset".
[13:26] Kooky Jetaime: The bytecode is fine
[13:26] Kooky Jetaime: the source is awol
[13:26] Adam Xinpeng: the 'sourcecode' which is what is not being found, and the 'bytecode' which is what's running.
[13:26] otakup0pe Neumann: yeah the bytecode is a different asset altogether.
[13:26] Zero Linden: well - the executing script is entirelly copied into the sim state
[13:26] Zero Linden: there is NO sharing there
[13:26] otakup0pe Neumann: the bytecode Zero ?
[13:26] otakup0pe Neumann: like, when it's running ?
[13:26] Adam Xinpeng: But the task inventory item will be the same among all instances since it would have been cloned from a single original.
[13:27] Zero Linden: so it can still execute even if we killed the data on the asset server
[13:27] Zero Linden: yes, the bytecode
[13:27] Zero Linden: and the state
[13:27] otakup0pe Neumann: yeah, state too. stack/ etc
[13:27] otakup0pe Neumann: That's interesting.
[13:27] Adam Xinpeng: Haha yes.
[13:27] otakup0pe Neumann: But if you de-rez it, presumably that would stop working.
[13:27] Kooky Jetaime: the inventory item yes but as I understand, every time something changes, a new asset is created
[13:27] otakup0pe Neumann: derez and re-rez'd
[13:27] Adam Xinpeng: LSL's bytecode is certainly interesting - you pretty much compile the whole application and it's runtime state together.
[13:27] otakup0pe Neumann: I'm still curious as to what can be done about these.
[13:27] Kooky Jetaime: so if you copy an asset, it creates a duplicate copy of the asset with a new UUID
[13:28] otakup0pe Neumann: Most of the time when I get script missing errors, the scripts show back up when the weather gets bad. But I've been getting gesture missing messages for about two months. Even when a stark naked ruth.
[13:28] Zero Linden: oN - no not at all
[13:28] Adam Xinpeng: Hrrm. Hazarding a guess here - but equipped gestures stay between sessions right?
[13:28] Zero Linden: if you derez it, the state, w/bytecode is all packaged up and stored
[13:28] Adam Xinpeng: It must be recording a list of what you have equipped - if that item was nixed, then that could do it?
[13:29] Adam Xinpeng: (that is it looks for it whenever you login because it's in your list of equipped gestures)
[13:29] Kooky Jetaime: Its stored and attached? to the assets uuid?
[13:29] Kooky Jetaime: or else how would it know to pull that stored bytecode out
[13:29] Zero Linden: Kooky - no, if you copy an item in inventory - that is just two inventory entries (rows in a DB table) that point to the same asset ID
[13:29] Zero Linden: When you load a script, the script has two assets
[13:29] otakup0pe Neumann: hmm. what about the bytecode/stack in that case ?
[13:30] Zero Linden: the compiled code and the source code
[13:30] Zero Linden: the compiled code is fetched
[13:30] otakup0pe Neumann: I don't know if we are drifting off topic but that doesn't bother me. Heh.
[13:30] Zero Linden: the bytecode loaded inot the 16k memory buffer for the script
[13:30] Zero Linden: the runtime state (stack and vars) are initialized
[13:30] Zero Linden: and the script is running
[13:30] Kooky Jetaime: So every time you copy something, your just creating another pointer in the database, right?
[13:30] otakup0pe Neumann: adam if i remember correctly i have tried this as a completely nude avatar, no attachmets, no animations, nothing
[13:30] Zero Linden: at that point, the running state, which includes the bytecode, has nothing to do with the asset
[13:30] Zero Linden: if you cause the state to be stored, derez, or sim state save
[13:31] Adam Xinpeng: otakup0pe, yeah but that doesnt mean the list isnt being followed around with your av?
[13:31] Zero Linden: then the whole VM state for that "script" is what is saved, bytecode, stack, vars and all
[13:31] Kooky Jetaime: And, say, in the case of notecards, if you modify that asset, a new asset is created and the database pointer is updated to point at the newly created asset, right?
[13:31] Adam Xinpeng: Yes - although I think notecards are 'special' in how they work.
[13:31] Adam Xinpeng: (that is they are an exception not the norm)
[13:31] Zero Linden: Kooky yes - that is how notecards work, or how scripts work when you edit the source
[13:32] Zero Linden: just not how stored script state works
[13:32] Kooky Jetaime: So are scripts.. but I figure notecards are simplier than scripts in there isn't two halfs of them.
[13:32] Zero Linden: I don't think notecards are handled differently - only in the UI
[13:32] Kooky Jetaime: Prims work the same way? Just a pointer to the original, until a change is made?
[13:33] Kooky Jetaime: or are they duplicated in their entirety?
[13:33] Adam Xinpeng: Prims are duped completely
[13:33] Zero Linden: duplicated
[13:33] Adam Xinpeng: You cant 'update' a prim set.
[13:33] Kooky Jetaime: ok
[13:33] Kooky Jetaime is making some serious notes here
[13:33] Zero Linden: once rez'd the link set / prim / script state of the object is a separate entity from any of the assets used to create it
[13:33] otakup0pe Neumann: even the root ?
[13:34] Kooky Jetaime: Well, I suddenly don't feel so guilty at making tons of copies of notecards or scripts.... its just adding to the database clutter not the asset servers themselves, until I change things.
[13:34] Zero Linden: The "store back..." menu options are just "syntatic sugar": the in world object remembers the inventory item it was rez'd from, and just "updates" the inventory item to point at a NEW asset that containst the "pickled" state
[13:34] Zero Linden: there have I mixed enough CS metaphors yet?
[13:34] otakup0pe Neumann: Almost. Heh.
[13:35] Benja Kepler: mmmm pickle
[13:35] Adam Xinpeng: We'll burn that bridge when we come to it. ;)
[13:35] otakup0pe Neumann: Touché on pickled though. That is universally understood.
[13:35] otakup0pe Neumann: I bet adam has questions about het-grid :s
[13:35] Adam Xinpeng: Yeah, that's my motherlode pack.
[13:35] Zero Linden: ask away
[13:35] Adam Xinpeng: <--- one of the lead OpenSim devs.
[13:36] Zero Linden: Yes, I know!
[13:36] Adam Xinpeng: We want to connect our sims up to the grid, what do we need to get ready for there? (We've got the support for plugin grid protocols already in)
[13:36] otakup0pe Neumann: OpenSim wants to be ready when the time comes.
[13:36] Adam Xinpeng: It's a year out, well aware of that - but I'd rather we didnt get locked into doing something stupid at so early a phase. Heh.
[13:36] Zero Linden: notes from internal IRC: Asset system now contains about 70TB
[13:36] otakup0pe Neumann: Woah
[13:36] otakup0pe Neumann: That's a lot of poseballs.
[13:37] Zha Ewry: Lot of pretty prim skirts
[13:37] Zero Linden: Okay - let's get terms straight
[13:38] Zero Linden: HetGrid is a near-term project (two months) that will allow LL to run different regions at different versions of the simulator
[13:38] Zero Linden: right now we only do that briely during a rolling update, and only for some kinds of changes
[13:38] Adam Xinpeng: I imagine all the infrastructure is still fairly 'trusting' though?
[13:38] Adam Xinpeng: IE: sims will still dictate L$ transfers, etc.
[13:38] Wyn Galbraith should clean out her inventory.
[13:38] Zha Ewry: But hetrgrid is still a monolithic, single asset server, single domain world
[13:38] Zero Linden: Het Grid will allow things like main grid beta regions, and new sim code being deployed over a few days (first 100 regions, then 1k, then all, say)
[13:38] Zero Linden: Zha, yes, it is.
[13:39] Zha Ewry nods
[13:39] Zero Linden: AX - yes, HetGrid is about deployment and version and code management
[13:39] otakup0pe Neumann: Zero this sounds like a precursor to mono.
[13:39] otakup0pe Neumann: among other things.
[13:39] Zha Ewry: Mono, Newer Havoc, big changes needing slow rollouts
[13:40] Zero Linden: oN - we need this before we can really release Mono
[13:40] otakup0pe Neumann: (i'd be happy to see a single sim dev grid to poke at mono with)
[13:40] Zero Linden: Mono will be the driving use case of HetGrid, though other things might end up using it first
[13:41] Zha Ewry: ON, you do realize that, Mono done right, at first, you won't be able to tell?
[13:41] Zero Linden: the first phase of HetGrid went to testing last week, and is being resubmitted today or tomorrow (there were bugs...!)
[13:41] Adam Xinpeng: Well, other than the performance increase.
[13:41] otakup0pe Neumann: Zha other than the speed boosts ? yeah
[13:41] otakup0pe Neumann: i realize.
[13:41] Zha Ewry nods
[13:41] Adam Xinpeng: LSL is so stupidly interpreted it's silly.
[13:41] Zha Ewry: Right. Perf boost, yes.
[13:41] Zha Ewry: But otherwise, no.
[13:41] otakup0pe Neumann nods
[13:42] otakup0pe Neumann: so back to the question though. what should opensim keep in mind to not code their way into a corner
[13:42] Zero Linden is biting his tongue
[13:42] otakup0pe Neumann: :s
[13:42] Zha Ewry: Just see a lot of people excietd about mono for all the wrong reasons.
[13:42] Zero Linden: Okay - the future grid architecture - with domains, as shown on the screen behind me
[13:42] Saijanai Kuhn: matricies? pointers?
[13:42] Saijanai Kuhn: OOP?
[13:42] Zero Linden: is what opensim needs to start being aware of
[13:42] Adam Xinpeng: Saijanai: that probably wont happen - not this year, probably not next year either. LL's going to be compiling the CIL on their systems.
[13:43] Zero Linden: AX / Sai - yes, initially, MONO is just a way to better execute LSL -- opening it up to other languages probably won't happen this year
[13:43] otakup0pe Neumann: i might be getting ahead... but what fits into each domain ? agent is everything not sim/(in world) assets/physics/script(execution vs compile) ?
[13:43] Adam Xinpeng: How much is moving into the agent domain - I'm assuming money transfers, maybe instant messaging, inventory, etc. What is the sim being authoritive in -- just display of the scenes contents?
[13:44] Adam Xinpeng: (and what about mixed cases, like asset downloads?)
[13:44] otakup0pe Neumann: do assets fit cleanly into one or the other domain ?
[13:44] Zero Linden: The split between Agent Domain and Region Domain is exactly the split between agents and regions --- so, thinks like groups and profiles are in the agent domain
[13:45] Zero Linden: things like parcel info, sim states, places, are in the region domain
[13:45] Zero Linden: L$ are difficult: av to av is purely agent domain
[13:45] Zha Ewry looks at Adam, and lifts an eyebrow. "You're begging the question, Adam, as to whether OpenSim wants to do both domains."
[13:45] Zero Linden: but paying objects in world with scripts involves both domains
[13:45] otakup0pe Neumann: Zha for now there isn't much choice.
[13:45] Saijanai Kuhn nots that long-term, mono and HetGrid are essential to a true "World Wide Web" of virtual worlds...
[13:45] Adam Xinpeng: Well, OpenSim already does both domains at the moment - we just dont want to get too stuck into the scaling side of grid things - madness lies this way & all.
[13:45] otakup0pe Neumann: So where do the poseballs fit in. Assets / Inventory ?
[13:46] otakup0pe Neumann: From what you are saying, it seems it straddles the line ?
[13:46] Zha Ewry nods "Well, right, both domains, down the road"
[13:46] otakup0pe Neumann: Depending on the asset ?
[13:46] Zero Linden: Assets are on both sides
[13:46] Zero Linden: !
[13:46] Saijanai Kuhn: scripting exists in both worlds
[13:46] Zero Linden listens for the hush
[13:46] Adam Xinpeng: Assets are going to be an interesting one - because we might want to have assets only the simulator has (not on the grid), and likewise we want to import assets from inventory and external content.
[13:46] Saijanai Kuhn: just about any asset has some presence in the cache
[13:46] otakup0pe Neumann: interesting. so i'm seeing multiple sources of assets then ?
[13:46] Zero Linden: yes, and asset use is a negotiated trust issue:
[13:47] otakup0pe Neumann: a linden telehub might come from LL's asset server, but a data visualisation script might come from a grad students desktop ?
[13:47] Zero Linden: Does this agent domain, that has the asset that is behind your inventory item, trust that region domain to respect the permissions required when rezzing the object?
[13:47] otakup0pe Neumann: (or some other arbitary location)
[13:47] Zero Linden: if not, the asset won't transfer
[13:47] otakup0pe Neumann: Interesting.
[13:47] Adam Xinpeng: Zero: I dont suppose you saw the email that came from a discussion I had on this with soft linden?
[13:48] Zero Linden: Vice versa - does this region domain trust that the compiled bytecode asset from that agent domain is okay to execute?
[13:48] Saijanai Kuhn thinks in the medium long-term about colloabratio between various P2P and client server entities not directly associated with LL
[13:48] Zha Ewry: Getting the trust right.. is one of the hardest issues in making a non monolothic multiple domain metaverse
[13:48] Zero Linden: (i know, I know, bytecode should always be safe....but that just isn't how LSL bytecode or CIL bytecode work.... sigh....)
[13:48] Adam Xinpeng: There's the inverse side of that too.
[13:49] Adam Xinpeng: You dont nessecarily let your websites visitors upload scripts onto your webserver - importing scripts onto your server isnt nessecarily something you want (although rejecting it easier than getting)
[13:49] Zero Linden: AX, no I didn't - please shoot me a copy: zero (dot) linden (at) sl (dot) com
[13:49] Zha Ewry: In the large, down the road, you'll have to assume that every part of the world may or may not be trustworthy
[13:49] Adam Xinpeng: Zero: I'll forward it over.
[13:49] Kooky Jetaime wonders if Zero is worried about email mining in world.. heh
[13:49] Adam Xinpeng: Sent
[13:50] Zero Linden: since I post the transcripts on the wiki - it is mining off the web
[13:50] Zha Ewry: LOL Yes.
[13:50] Kooky Jetaime: Ah. Ko
[13:50] Zero Linden: I'll look at it later when I'm in the office (I'm out of office at the moment)
[13:50] Adam Xinpeng: It's a pretty simple idea - but the gist of it was, encapsulating assets in a encrypted container, and controlling access to the keys, not the assets.
[13:51] Zha Ewry: Long term.. If you really want a ully open metaverse.. you have to validate trust between all the possible parts
[13:51] Zha Ewry: *fully
[13:51] Adam Xinpeng: That way, you could put abitrary rules on who gets what, and at the same time - not have to worry about proxies, or insecure servers having their disks trawled. (as long as it wasnt stored uncontained)
[13:51] Zero Linden: So the short of it is, this design, while initially will be all just LL serves, is being done expecting parts of it to be run elsewhere in some not-so-far off future
[13:51] Adam Xinpeng: I guess on that topic - long long term out - third party grids?
[13:52] Adam Xinpeng: Ideally the whole concept of a 'grid' dissapears - it's pretty abitrary.
[13:52] Saijanai Kuhn: 3rd party grids/P@P networks
[13:52] Saijanai Kuhn: P2P
[13:52] Zero Linden: We've had some discussions about encrypted assets before, here at LL, so yes, several different ways to handle 'em
[13:52] Zha Ewry nods Third parts services which can be combiend to creatas parts of the world
[13:52] Adam Xinpeng: Well, dont even need to go to that level - you can setup a website now without an authority. (although you may want one for DNS)
[13:52] Zero Linden: AX - I'm not so sure that "long long" isn't too pesemistic
[13:53] Zha Ewry: But you don't trust a web site as much as you need to trust a player in th gridd
[13:53] Adam Xinpeng: Why?
[13:53] Adam Xinpeng: Anonymous visitors to your site, SL is practically anonymous these days.
[13:53] Zero Linden: Sorry, there were WAY too many negatives in my statment
[13:53] Adam Xinpeng: If you want security, you use some kind of authentication.
[13:53] Adam Xinpeng: Hehe.
[13:53] otakup0pe Neumann: haha zero.
[13:53] Zero Linden: I meant to say, it won't be so "long long" from now
[13:54] Zha Ewry: Heh. I folllowed that Zero (Just nulled out all the negatives)
[13:54] Zero Linden: As for "3rd party grids" - I don't like the idea of "grids" - I think a fundimental reason SL works is because, unlike web sites, our builds are connected
[13:54] Saijanai Kuhn: I've already raised an issue on sldev and Croquet Dev about SL -Squeak client cooperation. Have you looked at things from a peer to peer POV instead of just a client-server POV?>
[13:54] Zha Ewry: I think, that to make something like a fully open metaverse, you need to validate trust (and expect man in the middle attacks) more than on the web for simple page fetches.
[13:54] Zero Linden: even private islands are more connected than private web sites are
[13:55] Zero Linden: after all, look at the kind of thing that happens around Spaceport Alpha, or Info Island
[13:55] Adam Xinpeng: Zero: that's been the direction we've been heading with OpenSim - the ideas we've been throwing around have been, using DNS, or encoding IP addresses into regionHandles (we are avoiding breaking client compatibility over that. heh.)
[13:55] otakup0pe Neumann: Zero what do you mean by that ?
[13:55] Ryozu Yamamoto: SLDNS would be region names, wouldn't it?
[13:55] Adam Xinpeng: The only things you need a "grid" for are users - you need a 'home server' for each user for things like IMs.
[13:55] otakup0pe Neumann: The organizational structure ? SciLands ?
[13:55] Zero Linden: So, I prefer that we build something that keeps the metaverse one grid in the sense that we all get to keep expanding this thing, even if there come to be
[13:55] Zero Linden: more parties running parts of it
[13:56] Saijanai Kuhn still wonders about P2P by third parities
[13:56] Saijanai Kuhn: parties*
[13:57] Adam Xinpeng: P2P isnt nessecary.
[13:57] Adam Xinpeng: Simple IP suffices.
[13:57] Saijanai Kuhn: ...
[13:57] Zero Linden: oN - I think that the kind of interaction that happens when the regions surrounding Spaceport Alpha are run by different entities, is just so much richer than if they all put links to each other on their web sites
[13:57] Zha Ewry: I think that, on the whole.. you won't be able to stop people from running totally isolated grids. But.. I think you want to make it as easy and desirable for third party services to stay affiliated
[13:57] otakup0pe Neumann: That's trueZha.
[13:57] Kooky Jetaime: You don't even need a home server for IMs.. you could easilly run a central IM server/system disconnected from the grid/any grids that allows for inter-grid communication...
[13:57] Zero Linden: there is something very integrating and symbiotic about continuous land, and being able to visit with one's persistent identity
[13:57] otakup0pe Neumann: One of my company projects =has= to be a private grid.
[13:57] Zero Linden: I heard this quote, which I love
[13:58] Zero Linden: "Without persistent identity, there is no community."
[13:58] otakup0pe Neumann: Due to the fact that it's a government project.
[13:58] Kooky Jetaime: yea, but if people like connectedness, why are there so many private islands :)
[13:58] Wyn Galbraith: Interesting, who said that?
[13:58] Adam Xinpeng: A single contiguous grid is going to lead to finite space issues though
[13:58] otakup0pe Neumann: And they have this thing about "secrecy".
[13:58] Ryozu Yamamoto: I imagine it would be very much like it is already, some islands seperate, some not.
[13:58] Zha Ewry: ON do you mean private in the sense of must be secure, or private, as in must not have *any* connections
[13:58] Kooky Jetaime: We are connected, even seperately by our choice to be here, Zero.
[13:58] Zero Linden: Wyn, I don't know, I heard it quoted at the Internet Identity Workshop, but I didn't catch the source
[13:58] otakup0pe Neumann: adam we'll have to stack them vertically
[13:58] Adam Xinpeng: That is, a similar problem to IP allocations today - unless you make the grid "architecture" like an oct/quadtree
[13:58] Wyn Galbraith wants an island just because it's an island.
[13:58] Adam Xinpeng: where you can infinitely subdivide the world.
[13:59] Saijanai Kuhn: P2P is a direction being take by at least some researchers. Eliminating P2P consideration from the "long view" eliminates cooperation between different kinds of virtual worlds
[13:59] Jansen Miles: Or an Internet/intranet paradigm. They both have their purposes.
[13:59] Wyn Galbraith googles it Zero.
[13:59] Zha Ewry: I want to be able to have an portion of the grid that is totallys sercurable, BUt.. I want to be able to walk off it, and into the real world, and back again
[13:59] Zero Linden: hmmmm "...infinitely subdivide the world..."
[13:59] Zero Linden: I think Zha has hit it upon the nose
[13:59] Adam Xinpeng: Zero: Basically think each "tile" can be divided into four, which can in turn be divided again.
[13:59] otakup0pe Neumann: Yeah Zero I agree
[13:59] otakup0pe Neumann: haha did the grid just close.
[13:59] Zero Linden: and, on that note, I'm going sign off... .thank you all for coming
[14:00] Zha Ewry: If I have to live in a bubble. I'm not happy.
[14:00] Kooky Jetaime: infinate subdivision isn't possible.. its easer to expand than contract
[14:00] otakup0pe Neumann: haha it did close.
[14:00] Zha Ewry: And. Zero. Thanks, as always for hosting
[14:00] otakup0pe Neumann: zero perfect timing ;)
[14:00] Kooky Jetaime: because sooner or later you reach the smallest possible part
[14:00] Adam Xinpeng: Not under an oct/quad tree
[14:00] Adam Xinpeng: you can just keep on adding nodes.
[14:00] Zero Linden: Lovely discussion - I'll try to get it up on the wiki soon
[14:00] Kooky Jetaime: Have a good day Zero.. thanks for the info
[14:00] Zero Linden: Thanks for coming, everyone
[14:00] Adam Xinpeng: Your finite limit is memory / complexity.
[14:00] Adam Xinpeng: Seeya Zero
[14:00] Zero Linden: Till next time (Thursday 7:30am...)
[14:00] Feep Larsson: Thanks, Zero!
[14:01] Wyn Galbraith: Thanks Zero.
[14:01] Zha Ewry: Now, mind you, to fully get it right, you need to make sure that stuff doesn't leak.
[14:01] Saijanai Kuhn: later