<LnL>
I'd like to have multiple builders to upload stuff but leaking the signing key is hard without making the rest super akward
<gchristensen>
I'm assuming you know how Hydra does it, with the hydra node doing all the signing locally
<joepie91>
LnL: standard setup is one Hydra instance that farms out work to the configured builders and then itself signs the resulting builds
<joepie91>
dunno how to do it without Hydra
<joepie91>
if at all possible
<srhb>
LnL: I've done both. I don't know how to do it without shipping at least everything through one builder that has the key, or exposing the key on every builder.
<srhb>
Alternatively, if you control the consumers too, just have a key per builder and add them all.
dermetfan has joined #nixos-chat
dermetfan has left #nixos-chat [#nixos-chat]
<gchristensen>
how about builders use individual keys, the leader trusts each key, accepts the key and resigns using the globallytrusted key and publishes
__Sander__ has quit [Quit: Konversation terminated!]
pie___ has joined #nixos-chat
pie__ has quit [Remote host closed the connection]
<LnL>
gchristensen: hmm, that's an interesting idea
<LnL>
a central build service would also be ok, but then I want a more flexible way to request builds
<infinisil>
Hashgraph is a decently new protocol with the same guarantees as blockchains but with 2 differences: It's *much* faster and efficient, it requires a closed set of participants (non-open)
<infinisil>
So if you see anybody thinking "I want to use blockchains, hype hype!" then you can ask them "Is the set of participants closed?" and if they answer yes you point them to this instead
<gchristensen>
if you have a closed set of participants, why not just use a database?
<infinisil>
A shared database? I guess it's not tamper-proof
<gchristensen>
I see
<infinisil>
__monty__: joepie91: ^^ (I think you participated in blockchains discussions previously)
<joepie91>
infinisil: their involvement in electronic voting makes me immediately skeptical
<joepie91>
and I can't help but feel like they're trying to profit off the *anti-blockchain* hype
<infinisil>
I don't care much about that company specifically, only the algorithm/data structure really
<joepie91>
and even that is overkill for most cases
<infinisil>
joepie91: Trying to profit of the anti-blockchain hype seems to be something good imo? blockchain is a waste of power, better to get away from it if you can
<joepie91>
multi-writer stuff is super rare
<joepie91>
infinisil: no, it's not, because it tends to involve hype in and of itself; exchanging one evil for another
<infinisil>
Ah thanks for that link
<__monty__>
Power consumption isn't inherent to blockchains though.
<infinisil>
joepie91: What's the disadvantage to chronicle? Just that it needs authorized clients or anything else too?
<__monty__>
Does it break down when the group of participants isn't closed?
<joepie91>
infinisil: compared to what?
<infinisil>
joepie91: blockchains I guess
<joepie91>
infinisil: then 'disadvantage' isn't really the right word, it's designed for a different usecase for which blockchains are currently (poorly) used
<infinisil>
joepie91: I'm thinking maybe it's not byazanite-fault tolerant
<joepie91>
it expects you to know the participants upfront, for example
<infinisil>
(which hashgraph is)
<joepie91>
infinisil: BFT isn't an inherently desirable thing
<infinisil>
Huh how not?
<joepie91>
same reason you don't want scope creep elsewhere; the bigger the scope / feature list, the more complexity is involved, which virtually always involves tradeoffs somewhere
<joepie91>
this is exactly the same reason why blockchains are unsuitable for most real-world usecases
<joepie91>
they just have particularly severe tradeoffs
<infinisil>
Yeah I guess
<joepie91>
the reason blockchains are so slow, for example, is that it tries to provide a rather extreme combination of guarantees under rather extreme untrusted circumstances
<joepie91>
if you don't have those extreme requirements, then you're just making an unnecessary performance tradeoff (and probably many other unnecessary tradeoffs)
<joepie91>
the tradeoffs aren't so pronounced for every technology, but the general rule holds true - don't use something with more complexity than you need, if you can avoid it
<infinisil>
Yeah, so as with everything else, if you don't need it, it only adds complexity
<joepie91>
yeah
<gchristensen>
sqlite, baby
<infinisil>
I'll look into hashgraph some more, it seems interesting. Couldn't find too much information on it unfortunately though
pie__ has joined #nixos-chat
<infinisil>
I don't like these fancy webpages that don't have a lot of content in the end :/
<joepie91>
infinisil: it's worth considering that there's a commercial company behind it - that is, one that seems to exist specifically for peddling that tech
<joepie91>
that's an automatic entry in the skepticism table to me :)
<gchristensen>
not sure that, specifically, should be a problem
<joepie91>
it is when there's no obvious valid business model
<gchristensen>
licensing the tech? :)
<joepie91>
and that right there is enough reason for me to lose all interest in it
<infinisil>
I'm also a bit skeptical, but I think maybe that's required if you want to content with the blockchain hype..
drakonis_ has joined #nixos-chat
<infinisil>
contend*
<joepie91>
and certainly puts it on the list of stuff I will never ever recommend for anything to anybody ever
<infinisil>
hashgraph is just an algorithm
<joepie91>
yes, so where's the business model
pie___ has quit [Ping timeout: 260 seconds]
<infinisil>
The thing they build on top of it I guess, hashgraph is just the base layer. Not entirely sure, but you don't need to use *their* implementation for anything, you can just profit from the algorithm itself
<gchristensen>
then you're not their target customer :)
<joepie91>
infinisil: unless it's patented
<joepie91>
and what's the thing they build on top of it?
<infinisil>
joepie91: Look at the pdf in the site i linked
<infinisil>
These swirlds guys have detailed technical reports even, with implementation details and stuff
<infinisil>
That's what I like to see
<joepie91>
"Swirlds, Inc. (Swirlds) is a Delaware corporation that is the owner and licensor of the hashgraph consensus algorithm. Swirlds is currently the sole member of Hedera Hashgraph, LLC, the expected legal entity for the Hedera Hashgraph Council. Prior to launch of the Hedera Hashgraph public ledger and the establishment of the Hedera Hashgraph Council, Swirlds will retain control of governance and network development, and Swirlds will be a permanent
<joepie91>
member of the Hedera Hashgraph Council. "
<infinisil>
They seem to have an (unfortunately non-open-source) SDK too
<joepie91>
that "if there's a company behind it for peddling the tech" heuristic isn't just paranoia :)
<joepie91>
I think I've probably seen a few hundred of these cases by now
<gchristensen>
depends if your problem is nicely solved by that tech and you have budget for it :)
<infinisil>
What the hell does a great algorithm use if nobody can use it :/
<joepie91>
generally ending with either highly sketchy business models (see eg. maidsafe), or patents, or the tech being BS and just a shiny trinket to sell to investors, or a play for a commercial monopoly over a particular type of software/service, ...
<joepie91>
the only cases of commercially-backed 'novel tech' I've seen that weren't super dubious, were those who were extremely upfront about their business model and their reasons for it
<joepie91>
this is why I automatically distrust things without a clear business model
<joepie91>
and this is a heuristic, not a universal rule, but certainly a reliable one so far...
<gchristensen>
QNever trust anything that can think for itself if you can't see where it keeps its brain.Q
<infinisil>
So this patent actually means that if I want to implement hashgraph I'll have to ask them and they'll probably say no?
<joepie91>
s/no/sure, pay us/
<gchristensen>
right^
<infinisil>
Algorithms shouldn't be marketable imo
<gchristensen>
that opens a big thing
<infinisil>
Imagine binary search were patented lol
<joepie91>
infinisil: see also: why software patents are so reviled :)
<joepie91>
but yeah, my POV on tech is pretty much "if it's not in the public commons, it might as well not exist"
<infinisil>
I'll probably have to agree on that
<gchristensen>
shame you can't use any software from the US since there is no provision is US law for things to enter the public domain ;)
<joepie91>
gchristensen: I'm defining 'public commons' a little more broadly here than 'public domain'
<gchristensen>
at best you can issue a license for a long time, but it isn't even possible to issue a "forever" license: they all expire after a while
<joepie91>
today in vendor tales: ask vendor months ago about a feature they claimed to have but that seemed unimplemented, never get a response, see a release announcement today announcing that they've added that feature in the new release...
<infinisil>
"Ticket closed. Reason: User incompetence, feature was there all along"
<gchristensen>
nice :)
<joepie91>
lol
<joepie91>
infinisil: not quite that level of shittiness
<joepie91>
but they do have a habit of just... stopping replying, when the questions get too icky
<joepie91>
it's a good thing the code is open-source :P so I can just fix it myself