Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
lgierth
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
42 ms
·
151.
▲
by
lgierth
9y ago
IPFS can discover other nodes in the same local network via mDNS -- you don't need to have an internet connection at all to share data locally. Combine this with the recent work on PubSub [1] and CRDTs [2] and you can make many applica
152.
▲
by
lgierth
9y ago
Nice to see this has made it through the standards process -- the `immutable` keyword is tremendously useful for systems that store and provide actual immutable data, e.g. content-addressed distributed systems.
153.
▲
by
lgierth
9y ago
And Vultr gives you BGP Anycast :)
154.
▲
by
lgierth
9y ago
Are whoppers a scarce resource?
155.
▲
by
lgierth
9y ago
I've had to deal with Hetzner and IPFS too -- my conclusion is that it's Hetzner who are aggressive here. In of the cases I had fixed the dialing-local-networks behaviour, and then Hetzner still continued to block the server for a
156.
▲
by
lgierth
9y ago
Very interesting! We've been using HAMTs in IPFS [1] to make huge directories more efficient (think NPM or Wikipedia), and the memory profile has been a pain, so this looks like a welcome improvement. [1] https://github.com&
157.
▲
by
lgierth
9y ago
Sorry I forgot, here's another paper laying out the Proof-of-Replication and Proof-of-Spacetime: https://filecoin.io/proof-of-replication.pdf
158.
▲
by
lgierth
9y ago
There's an alpha implementation of the IPFS and libp2p stack here: https://github.com/ipfs/js-ipfs and https://github.com/libp2p/js-libp2p
159.
▲
by
lgierth
9y ago
> SoundCloud just chose to not solve it once it became too hard on their database, including when they famously had issues with being early adopters of Mongo Do you have more info on the Mongo angle? I worked in SoundCloud engineering f
160.
▲
by
lgierth
9y ago
Because you need to keep proving over time
161.
▲
IPFS DNS Support
(blog.neocities.org)
2 points
by
lgierth
9y ago
|
0 comments
162.
▲
by
lgierth
9y ago
The idea of domain fronting though is that the fronting website is "expensive" to block that the censor won't do it. As with everything else, this isn't a 100% technical solution, it has a political component of how expe
163.
▲
by
lgierth
9y ago
Ah nice, I didn't know meek does domain fronting too, and thanks for mentioning snowflake.
164.
▲
by
lgierth
9y ago
> VPNs can also simply protect against attacks on public networks in e.g. cafes, libraries, airports. To expand on that, you could argue this for any network, even home and company networks -- who knows which of the many windows computer
165.
▲
by
lgierth
9y ago
We need something that tunnels over HTTPS/Websockets, and utilizes domain fronting [1] with e.g. Google Cloud or Azure. Ideally it'd also work over WebRTC in unreliable mode, but I'm not sure that works with domain fronting.
166.
▲
by
lgierth
9y ago
> VPN services can't exactly hide behind being an impartial carrier since they sell themselves as tools precisely for circumventing censorship, tracking, region locking, and providing anonymity. I really wish some VPN providers woul
167.
▲
by
lgierth
9y ago
It looks more like the various coins have been going down all day today?
168.
▲
by
lgierth
9y ago
I've been using schleuder through various mailing lists over the past years, it's always been doing what it says on the tin :)
169.
▲
by
lgierth
9y ago
> The benefit of this approach over the alternative (making people encrypt and sign with a lot of keys) is that once received, the content does not give away the key names of the other mailing list members. The other benefit is that memb
170.
▲
by
lgierth
9y ago
Filecoin and IPFS assemble the world’s knowledge and data into one big content-addressed graph (Merkle DAG), and attempt to make it accessible to any device, anywhere. This is great because it allows for all kinds of use cases apart from ju
171.
▲
by
lgierth
9y ago
Token sale is gonna start on the 27th, here's more info: https://protocol.ai/blog/ann-filecoin-token-sale-and-new-pap...
172.
▲
by
lgierth
9y ago
Filecoin and its underlying protocol IPFS assemble the world’s knowledge and data into one big content-addressed graph (Merkle DAG), and attempt to make it accessible to any device, anywhere. This is great because it allows for all kinds of
173.
▲
by
lgierth
9y ago
Apparantly FCC had to throttle their commenting system. > IMPORTANT NOTE: these numbers STILL represent only a portion of the final totals, and due to the massive numbers, comments and emails will be delivered over several days. https:&
174.
▲
by
lgierth
9y ago
Yeah I feel you, the management overhead is a big pain for many people at the moment, ourselves included. There's a couple of improvements to this coming in the next two months: 1) a new transport (QUIC) which has less overhead, 2) sma
175.
▲
by
lgierth
9y ago
You won't have to sit on the retrieval path with Filecoin. Nodes participating in the retrieval market will probably simply expose a Filecoin-enabled version of the HTTP-to-IPFS gateway. The v2 paper will have more details very soon. I
176.
▲
by
lgierth
9y ago
There are a ton of technology differences which the upcoming Filecoin v2 paper will address (very soon!). On the conceptual layer there's a difference in that Sia and Storj seem to be a bit more geared towards consumers and file storag
177.
▲
by
lgierth
9y ago
I hadn't looked at the planned changes for go1.9 until now and I have to say I'm really happy :) - net.Resolver.Dial allows doing all DNS resolution myself, right in the same process (via a localhost port). And nice it can do both
178.
▲
by
lgierth
9y ago
Sublime Text does this in a limited way by opening up in the state you left it, i.e. including unsaved changes, pane layout modifications, etc.
179.
▲
by
lgierth
9y ago
The correct dataset (with gzip compression) is here: https://ipfs.io/ipfs/QmcvfB6pAqUfTnuAK8zFKVxbdhopnBPveJrDcy1...
180.
▲
by
lgierth
9y ago
Oops, it turns out I somehow ended up with decompressed files, that's why the filesizes are so far off. I'm creating a new fixed dump.
More ›