#dat

Hey, rabin, or content defined chunking, is kind of a hash function that finds the boundaries by itself right? I was just wondering why people don't use it for locality sensitive hashing, where people often seem to do hashes of fixed size chunks. I was just thinking that cdc would give better results when looking for similar patterns, or are the chunks too

big in cdc algorithms? Does any of what I'm saying even make sense?

(somebody1234) mafintosh: is it possible to expose the server peer object received when connecting to the server in peer-network?

eljee: well the dat-design bump didn't seem to green travis: https://travis-ci.org/datproject/dat.land/jobs/...

ogd: eljee and i talked on friday about swarms in dat.land, curious if you want to jump on a call to hear your ideas on the best way to architect it. mafintosh mentioned having a websocket gateway last time we talked, curious if anyone is still working on that

@somebody1234 sounds like a good idea

@somebody1234 make an issue and i'll get to it

also, morning everyone

(Kriesse) Morning everyone :)

(Kriesse) @eljee @clkao Travis not greening on the dat-design bump is on me, fixing it now.

Thanks @kriesse

juliangruber: ping

juliangruber: can you do a new build of dat-desktop? i have my gov dataset sharing working now

i wanna send them a link to the desktop client

jhand: we need to fix some of the ux bugs when resuming live dats

kinda buggy right now in the output

mafintosh: nice

karissa: https://github.com/mafintosh/dat-virk-dk

mafintosh: sweet!

i'm gonna hack on peer-network a bit now

eljee: sort of what we talked about friday? https://github.com/datproject/dat.land/issues/215

karissa: yep

eljee: i just got server rendering the file list to work

karissa: which method?

eljee: ill open a pr

karissa: ok. can you also make it clear any code paths that are "experimental"

eljee: got it. i removed the webrtc thing

eljee: in the backend

karissa: ok

karissa: i was going to comment out webrtc on server side in `router.js` and just make it execute on a ?webrtc=true or some param like that

karissa: never got around to it on fri

aye

that might be smarter than what i did which was just nuke it

up to you

karissa: dont nuke all the way, just parameterize so we can experiment (if it gets better)

mafintosh: i'm making a new build right now

eljee: https://github.com/datproject/dat.land/pull/217

lunch time now

mafintosh: i want to work on a dat based auto updater for electron next, the current solutions all require a server to be in place

shouldn't be too hard, just launch a local http server that serves the dmgs via hyperdrive-http

mafintosh: https://github.com/juliangruber/dat-desktop/rel...

juliangruber: sweeet

downloading now

juliangruber: how difficult would it be to cut a windows release?

mafintosh: i haven't run it once on windows yet, i'll try now

first need to set up a vm though

hopefully i just need to make some path strings os aware

utp might not compile

i set up a windows box this weekend for that actually

unsure how the build setup for recent is now though on windows

mafintosh: which windows version would be most relevant to test on?

10 i guess?

i have 8.1 on my build box though

(i'm afraid)

i'll try 10

haha yes Windows is scary

juliangruber: usually windows is backwards compat

juliangruber: https://github.com/mafintosh/utp-native/issues/8 - did you try and compile on windows?

i'm just curious if it actually compiles hehe

mafintosh: not yet, first compiling issues for what's missing then trying to fix them one by one

weeet

mafintosh: can you enable issues on https://github.com/mafintosh/node-sodium-prebuilt?

juliangruber: done

so i am relooking at supporting rtc in datland server side (via haus). it seems we need a new hybrid-swarm wrapper capable of handling multiple discovery keys, ie the join semantic for rtc-based swarms.

clkao: I am not sure we want to do webrtc in the server side

we don't? then if you create a new archive in browser A, and browser B hits server it will not get server rendered results

clkao: I think it's quite brittle. mafintosh was mentioning it might be better to have a websocket based gateway

ok... i thought it was going to be another option

clkao: browser based dats are pretty ephemeral and I think it might be nice to prioritize or push ppl to use the desktop app and cli to keep dats up and running

clkao: what do you think? What does a user get from server rendering a browser shared dat?

clkao: like vs. just using web rtc

clkao: in the browser

i think the transport shouldn't matter. once we have a hosted service with api to replicate a given dat, the browser can well make a request to make the data sycned over, and close when it's indicated as replicated

clkao: yea, transports shouldn't matter. all you need is a stream

mafintosh: the problem i have now is that discovery swarm has the join interface and semantics, whiist rtc swarm has its discovery key baked into signalhub(swarmKey)

true. can't you just make a swarm per key?

that's what i was saying (for sanity check ;): a new hybrid-swarm wrapper capable of handling multiple discovery keys

ah cool

sry read half of o/ hehe

we can already tweak the underlying discovery-swarm of the hybrid-swarm to join for additional keys

and that's how server/haus.js works

Sure makes sense

I guess I just mean we should prioritize non webrtc because the other connections are better

*nod* do we have a socket-based swarm module already?

karissa: in any case we'll need the hybrid swarm interface to support multiple keys, once we have the scoket-base swarm in the family ;)

Makes sense!!

good morning party people

mafintosh: for discovery-swarm, if you join multiple keys, you have peers in the same pool right? as in not all peers you connect to will have keys they are you interested in

clkao: yup

up for the user to figure that out

normally the first message you send over the connection describes what you are interested in

and then you'll filter out stuff that way

(we do that in dat as well)

gotcha

pfrazee: do you envision beaker as an app people will have running more or less the entire time they have their computer open?

like a normal browser today

mafintosh: yeah

nice.