#dat

(lukeburns) @e-e-e looking at hyperchat -- pretty cool! are you imagining a situation where a participant's peers might not have the ability to read each other's feeds? seeing that `heard` appends discoveryKeys rather keys.

(e-e-e) @lukeburns yeah - I was thinking that if a peer was not connected to another in the chat history they should not be able to reconstruct that history.

(e-e-e) its really basic at the moment, the toughest part is how to share keys between peers that join - as it would be nice if someone listens to you, you could automatically also listen back. but at the moment it means you all need to share keys in some other way.

(lukeburns) @e-e-e one step in that direction might be to auto-connect to keys that connected peers have heard. e.g. if alice and bob are listening to each other, bob and charlie are listening to each other, and bob announces that he heard charlie and alice via their public keys, alice and charlie can listen to one another.

(e-e-e) interesting - perhaps as a config option it would be nice - as one way it keeps the conversation more closed, the other openning out. If you have one public key you can get them all.

(lukeburns) this doesn't feel too strange to me -- usually when i join a chat room i can see everyone who's in it. it's a little bit awkward if alice is receiving partial information about bob's communications with charlie (when messages are heard).

(lukeburns) but i could imagine there being scenarios where you might want to use more caution

(e-e-e) I am really open to contributions - I started playing with it as a way of getting more familiar with how hypercore works - but also see lots of ways it could be extended out. Eg. Defining a clear protocol rather than json for storing the data, option for displaying history etc.

(lukeburns) cool! i've been playing around with it a bit and ran into that question ^

(e-e-e) totally agree with you - not odd for chat room scenario.

(e-e-e) :)

(e-e-e) I need to also expose port options, as its blocked for me at work which is annoying

(G-Ray) @e-e-e : I wrote a basic hyperchat script 1 year ago :P. You can exchange peers when a new peers is introcued https://github.com/G-Ray/hyperchat

(e-e-e) @G-Ray that’s hilarious - thanks for the tip and sharing.

(e-e-e) i will have a look at how you handle it.

(G-Ray) Yeah but hypercore'api changed quite a bit since then I think

(e-e-e) yeah looks like hypecore-protocol has changed a fair bit - but could probably use custom userdata. Did not even think it was a possibility, so thank you.

(e-e-e) @G-Ray just got a PoC working with the new protocol - thanks for the hot tip. https://github.com/e-e-e/hyperchat/commit/d8785...

@e-e-e you can also send custom extension messages over the protocol now

(e-e-e) mafintosh: cool, is that documented?

ehhhhhhh

prob not haha

(e-e-e) ha ha

but it should be!

(e-e-e) no probs. I will dig in the source to see what you actually mean by extension

(e-e-e) ah sweet

(e-e-e) that is super interesting

(e-e-e) so basically all metadata info that i dont neccessarily want to store in the append only log, I can communicate via building an extension.

great !

yea!

simple dat test/debug tool, https://github.com/mafintosh/dat-test

I would still like to create a dcentralized storage space like freenet. I can't decide if I should choose dat or libtorrent (because every files should be encrypted and should be splitted into chunks of same size I guess)

Any hints ?

everyone - chjh_ does cool science stuff and is looking to learn JS - what are people's favourite resources?

blahah: nodeschool.io ?

in particular, to avoid going down the 2017 JS frameworks road to hell, just straight up clean JS

G-Ray: +1

G-Ray: bittorrent isnt great content that updates

mafintosh: Yes but if a content is encrypted, it can't really be updated efficiently right ?

A user would need the dat hash + a private key to decrypt the content

blahah: JS for cats

You still need a mechanism to push out new updates

And verify them etc

mafintosh: yeah right

yoshuawuyts: +1

and there's https://github.com/yoshuawuyts/tiny-guide-to-no...

but that's for once you've got the basics

mafintosh: How is calculated a block size in Hypercore/hyperdrive ?

G-Ray: its variable block sized

mafintosh: can it be set to be static ? (if content is a blob/ encrypted, what is the block size ?)

also chjh_ save these by substack (http://substack.net/how_I_write_modules) and mafintosh (http://mafintosh.com/pragmatic-modularity.html) for later

Thanks everyone for the suggestions!

(lukeburns) mafintosh: is hypercore-protocol handshake data (id, live, userData) encrypted?

@lukeburns yea! everything is

(lukeburns) thanks!

hmmmm

so i want to make a browser thing that uses dat (or just hypercore, really)

cblgh: yea?

i want to basically serve a index.html, or allow people to download it and serve themselves

could i use browserify to accomplish that?

cblgh: you have to use the ram or indexeddb storage provider

oh cool, so that's what the db in the example refers to?

cblgh: ah, thats a bit out of date sorry. karissa, jhand: how do i update that?

cblgh: https://github.com/mafintosh/hypercore

cblgh: pass in random-access-memory as the storage arguemtn

ah rad

and then hook up feed.replicate() with a websocket-stream or webrtc-swarm stream

just to make sure i'm on the right tracks, i kind of want to do a k,v store kind of thing

what kind of data is the values? files?

would hypercore be a good choice, or should i be looking at something else instead?

basically just bools

nothing big, just want to be able to ask for a location and see what the corresponding value would be, if there is one there

hypercore is an append-only log so all keys are incrementing numbers basically

we have a basic kv store on top called append-tree

you can pass in a hypercore to that

and then i'm working on a new one here, https://github.com/mafintosh/hyperdb

ohh nice, i think this is more like what i'm looking for

i was going to go with orbitdb for fun

the last one is exactly what you want but that is stil under developement

but looks like haad's been out of commission for a while

oh too bad, thats a fun project :)

yeah, i agree

distributed kv stores are trickier to build that you'd expect

how so?

consistency issues?

lots of the literatture evolve around making distributed kv stores for a data center

and not over the whole internet :)

so lots of new research

ahh

i've skimmed amazon's dynamo paper a bit

yea thats a great paper

btw will tr.put update a value if the name already exists with a different value?

cblgh: i have a fun little gist here, https://gist.github.com/mafintosh/0abb544b8b474... for hyperdb

cblgh: yes it will. be aware that append-tree is single writer though so only one peer can update it

hyperdb is multi writer

but again, super alpha

yeah, i'm not doing anything production so no worries

if i used hyperdb like in the gist, what kind of issues might i run into?

just random tihngs not working :)

i basically just want to read & update x, y positions

hahah

its surprisingly stable actually right now

but no guarantees yet

the interface is stable as well

oh rad

i'm releasing iterators soon

right not it is strictly put/get - iterators make .createReadStream possible

*now

:3

i'll try some things out and see where i end up, thanks for the info!

cblgh: the gist is kinda fun. the db is 500k of english words

woah what

so you run it with a word to check if it is english

i'll try the gist

try looking up the same word twice

on 2nd lookup its *much* faster cause its replicated

brb

203ms first time