#dat

@serapath ofc I gotta suggest beaker https://beakerbrowser.com

(serapath) what do you mean? because it supports dat natively?

yeah

and it's made by a handsome devil

(serapath) actually thats pretty awesome - but most ppl use mainstream browsers and its a requirement that it works on them

yeah

we gotta get a standards track started!

(serapath) i'm actually working on the web IDE of ethereum where they allow coding "solidity contracts" ...i'm not too interested in solidity and ethereum for now but i'm learning a lot and it pays something :-)

(serapath) yes please - standards track sounds awesome :-) i plan to slowly switch away from chrome and maybe either use beaker or brave, but ideally it would be easy to fork and put together my own browser

I understand that. Lots of money in the ole chains rn

relevant link I just clicked: http://continuations.com/post/161091248805/some...

(serapath) i'm very critical towards existing crypto currencies anyway. I'm not sure if the future can or will do without crypto currencies, but the nature of all existing crypto currencies i am aware of seems to be broken to me. ...but other than that i try to make a living and for now i had the luck to get a deal to work a few hours every week remote when and from where i want - which brings little money, but enough to keep lear

(serapath) i wish i was a faster and better programmer, but i still feel like i'm improving even though i wish i would learn faster :P

yeah I think that's great. Plus there is a lot of talent, and interesting ideas, in that space. It's a good place to be

just get your payment in fiat :D

(serapath) yes i do - it seems not practical to receive it in bitcoin or ether for now.

(serapath) do you think in the long term future there will be crypto currencies around?

I'm personally very skeptical of proof of work. If ethereum cracks proof of stake, then I'd have more faith. Otherwise, I think the solution will be a better payments messaging network; something similar to the interledger protocol

but the efficiency & governance problems just dont add up to me. Throughput is bad, you have to ship around a lot of extra data, the energy costs are high, governance is defacto plutocratic with mining power == votes on protocol decisions... it seems like a lot of extra pain for meager gains

(serapath) i'm sadly not familiar with the details of that stuff. i have some opinions and knowledge about currencies and economics in general, but that's why i'm trying to learn more about crypto and p2p in order to one day be able to judge the differences between all those systems and compare them with the knowledge that i have from my past

I'm a huge believer in service decentralization, but I think network-wide strict consensus is not a requirement to accomplish that, so I'm much more interested in Dat

yeah

(serapath) i find i really hard to judge :/

(serapath) yes

it's complicated for sure

Can we install Dat on Windows now?

HID_System: let me check, jhand karissa ^ ?

HID_System: if you have npm/node, you can check pretty quickly with `npm i -g dat`

Yes, but what about native moudles?

HID_System: I dont have a windows box so I dont know off the top of my head, but if installing dat works, either it's because the native modules are working, or it's because dat has JS fallbacks

pfrazee: Windows builds still failing.

:\ ok

sorry, guess they havent gotten there yet

HID_System: utp isn't required so may still work

We were seeing some windows bugs haven't had time to debug though

But a successful build is better.

Def

The build passed.

(serapath) another problem

(serapath) we did not start the syncing process yet

(serapath) ```bash

(serapath) dat ninasbackup/

(serapath) dat v13.5.0

(serapath) [full message: https://gitter.im/datproject/discussions?at=592...]

(serapath) maybe the problem is to type `dat ninasbackup/` instead of `dat share ninasbackup/`

(serapath) anyway, now i tried: `dat share ninasbackup/` to start it again, but now it says: (empty archive)

(serapath) ```bash

(serapath) dat ninasbackup/

(serapath) dat v13.5.0

(serapath) [full message: https://gitter.im/datproject/discussions?at=592...]

(serapath) ...hmm, now it's doing something. i just started it with `dat share` and it starts again counting files and megabytes

wa7son: thanks

@serapath didnt you say it's like 13gb? You might try a smaller dataset (or a subset) until you're sure things are working

mafintosh: i've been trying to wrap my head around how to make a custom storage for a hyperdrive that would actually be selections from one or more other HDs (i think this is what you suggested to blahah a few days ago). i assume it would have its own key, but it would somewhere need a reference to the larger, original HD's key (to be able to join the swarm)

-- how this would work for multiple other HDs i have no idea. but i think this is probably not correct because then i'd be implementing a connection or discovery inside of storage... so i guess basic question is at a high level, how did you imagine doing this?

also, I've looked at poga's hyperdrive-ln, which is nice but not ideal here because i want these smaller hyperdrives to be discovered as sources for the limited content that they do have of the larger hyperdrives.

pfrazee: i see in documentation that BB automatically renders markdown; out of curiosity is there any plan for a markdown editor? (just tried to check if there was already one, but can't create a new site (did file an issue tho) so asking here instead)

aaaaaaaaa____: I'll check into the issue. Would definitely be easily to make a markdown editor in userland

aaaaaaaaa____: replied to your issue

pfrazee: ah i see, hadn't seen this before! https://beakerbrowser.com/docs/tutorials/write-... .. up till now i had only seen opening the folder as the only thing in the interface pointing to how to modify the dat, but this is great

aaaaaaaaa____: Im signing off for the night but ping me later if you need any help

i will but this looks pretty straightforward for what i want to do (edit a markdown file in a dat without opening the folder, finding the file, and opening an editor)

mafintosh: for browser support, is deriving the key pair necessary? could the public key be saved when the private key is saved instead?

inside of hyperdrive

substack: yea that already works

substack: just derive your own signing keypair

Pass in the secretKey and set storeSecretKey to false

with a blake2b implementation this could be done upstream in sodium-javascript https://download.libsodium.org/doc/key_derivati...

scroll down a bit, there are two anchors with the same id

emilbayes: haha I just saw your blake2b module

published 1 hour ago

you must be going down the same path as I've been going down, I will work on something else for a few

substack: Hehe, I have the code almost ready

have ~600 test vectors

512 passing

substack: Needed the salt and personal buffers to be in the parameter_block, something which only libsodium has

because it's not part of the spec, but it's in the BLAKE book

(and paper)

excellent, thanks so much for working on this!

hehe

I've been thinking how to implement hyperlog on top of hyperdrive directly too, have some good ideas

substack: I have to get ROI on buying the blake book

substack: mafintosh and I are working on a multiwriter database on top of hypercores, has he talked about it? It reminds me a lot of hyperlog

though the links are not explicit as in hyperlog

are the links content-addressed?

indirectly

I like the idempotent writes using content-addressed storage

you can trivially implement a content addressed dag on top of this new thing

nice

it's a causal linker basically

which doens't say much

I just don't want to duplicate hashing if the underlying layer already does it

to save some cpu cycles

you wouldn't need to i think

cause we already hash the value directly

another thing I would like in hyperlog is if you could point at hashes you haven't seen

thats what this new thing allows you to!!

:D

nice

basically the main feature

other than that just a hyperlog ish thing

do you think it would be possible to present the existing hyperlog api on top of this new thing?

plan on writing a white paper about this new thing

yea!

for the most part, things like .replicate() don't have to be implemented if hyperdrive will handle that

on the top of my head there should not be any tricky parts in getting that to work

basically our new thing is just a kv store

supporting concurrent writes

with history support

and sparse lookups (i.e. point to values you havent seen yet)

all that adds up to hyperlog

do you think it makes sense to store other things like the list of heads in hypercore or better to use another storage layer?

and same question but for higher-level materialized views

like building hyperkv

although maybe a hyperkv wouldn't be necessary if you already have a key/value store? I would have to look more closely at this thing to know for sure

we try to store everytging in hypercores

to get free replication

another question: how does this k/v store handle replication and multiple values for a given key?

api is key -> values

ok

so up to user to reduce

test vectors for the curious: https://github.com/emilbayes/blake2b

like hyperkv

with a kappa architecture, you might not want to replicate the materialized views

because this way, clients can be running different versions of the indexes

or be running custom indexes

but they can still replicate

yea

this is kinda like a distributed index

substack: that's exactly what I want too for my use cases

also for space reasons you might not want to store the history of materialized view writes

only the most current snapshot, which can be derived from the history

so with our index you'll only end up storing the latest values

unless you replicate all of the history

but! if you want to build a historical database with historical indexes, you could also store the materialized views in hypercore which would be very cool

time traveling database

substack: thats exactly what we are aiming for

so its not going to replace kappa arch

so I can envision that sometimes you would want to do this but not all the time

more like an addition to it

with hypercore is it possible to sparsely partition where data is stored and also possibly delete it (really delete it)?

like for example if you have a materialized view in hypercore but then a new update comes along where you need the materialized view to work differently and want to free up that space