#dat

mafintosh: why is hypercore-protocol replication capped at 128 feeds?

urgent question: if "dat boi" were legit middleware for the dat network, what would the initials BOI stand for?

lritter: latency optimised infrastructure

mafintosh, that's loi.

noffle: arbitrary, has be capped somewhere

lritter: oh right lol, early

lritter: bits over infrastructure

hhah

mafintosh, mh. well it's something.

mafintosh, Blockchain Offline Import

Bidirectionally Ordered Index

Bootstrap Over Internet?

lritter: the birectional one is spot on

that's the one then

@RangerMauve https://github.com/mafintosh/dht-rpc/pull/2 <-- doing an iteration of the base dht here

will merge in a few hours. next step is implementing some of the security features we've been talking aoubt (main focus first is getting the api right an scaling)

cool, the freedombone project uses dat for distribution https://freedombone.net/install_desktop.html

ralphtheninja: nice

lritter: beaker ouster identity

edsu, noted :)

a question about the dat protocol: how's dat performing when it comes to the classic strengths of bittorent, i.e. transferring huge files, or just small chunks within big files?

i'm particularly thinking of designing a p2p protocol for a decentralized VR space

people need to be able to share assets with each other, and some of those could be larger binary blobs -- or could they? i guess that's what i wonder about

lritter: dat is only okay at that right now but we've got improvements rolling out. I think it's a good bet in the long term

maf has done a bunch of work on indexing structures and gotten really good benches out of it

pfrazee, nice

pfrazee, are you aware of a high frequency p2p protocol right now that could be used to share lossy realtime data like 3d orientation and such?

i'm thinking of something UDP based

lritter: it depends on some of the properties you're looking for. What do you need in addition to being able to establish UDP channels between all users in the same area?

e.g. do you need it to work without having complete connectivity between all nodes? Then you probably want to look into gossip systems

pfrazee, the idea is that a protocol like dat would be used to share static assets and other low freq data that doesn't change much, and the hifreq protocol would be used to share realtime events; i don't know if it's really necessary to manage a session tho

i'd love something like region based discovery where ip's start talking to each other because they are spatially close in VR

lritter: yeah that makes sense. I haven't researched the realtime channel question so I can't give great answers. I'm sure there's some work being done because of projects like decentraland. Of course, the games world is going to be a big source of past work, but their requirements are specific

i guess there would be a tcp based midfreq protocol for handshakes and chat, a udp based hifreq protocol for realtime events, and dat for the assets

pfrazee, yeah i'm mostly based in games

i'm dreaming of a VR analogy to the web for 20 years

yeah for sure

if it was possible to get that working without a central matchmaker or commercial intermediary, that would be amazing

off the top of my head, yeah I'd probably use a discovery system similar to what we're building with hyperswarm, but modified to lookup peers based on "regions" as opposed to "content" topics

think "second life" but without linden labs :)

for sure

yeah

pfrazee, i guess i need to read myself into this topic of discovery

lritter: that'd be the first step because that's how you're going to arrange connections between virtually-colocated users. After that you need to figure out your requirements for the network after some connections are made

pfrazee, i guess one could just repurpose topics to mean "rooms"; that would be specific enough i think

right

pfrazee, did you know that most of GTA V's online services are p2p based? one user is always hosting the session for everyone else

I'd bet that your main challenges will be 1) counteracting delays to get a connection, 2) counteracting peers in the same region that cant establish connections to everyone else in the region, 3) latency once the connections are established, and 4) malicious behavior

lritter: yeah that's interesting. It means 1 user always has better latency than the others, and Im sure they had to work really hard to counteract cheating

pfrazee, they completely fail on the cheating problem

yeah that's not surprising

for VR, those 2 issues may actually be okay

for an unconstrained space, there are less ways in which things can go wrong

but as soon as constraints come into play, it gets difficult

like, whoever runs the session controls some of the rules, sure, but you *can* just work that into the system

you could set it up where regions simply have owners and they're expected to set those rules

only thing i can think of is that each client validates the data it's seeing and penalizes non-cooperative peers

plus if those owners go offline, the region goes down

you could do that too, but without appointed authority, it becomes he-said/she-said

which means the participation will segment

it's a tricky problem.

godamnit i always miss the good conversations

except, geographically close is not the same as low-latency between (network-close)

in the end you really dont even need dat for this kind of vr space

you could easily do it with portals to other nodes ,

people host there game , people share there nodes ip addy , yes the person who started the game has supreme power but they will either play nice and not use there admin powers or abuse them and people will just not participate on that server

if anything dat could be useful for a discovery program

not sure how good dat is punching through nats

if it could do something there that would also be helpful i suppose

if you want something that is seamless across nodes, just space people apart based on latency distance

turn space into a loading screen