stebalien changed the topic of #ipfs to: Heads Up: To talk, you need to register your nick! Announcements: go-ipfs 0.4.23 and js-ipfs 0.41 are out! Get them from dist.ipfs.io and npm respectively! | Also: #libp2p #ipfs-cluster #filecoin #ipfs-dev | IPFS, the InterPlanetary FileSystem: https://github.com/ipfs/ipfs | Logs: https://view.matrix.org/room/!yhqiEdqNjyPbxtUjzm:matrix.org/ | Forums: https://discuss.ipfs.io | Code of
MDude has joined #ipfs
chrispeel[m] has joined #ipfs
cipres has quit [Ping timeout: 250 seconds]
D_ has quit [Quit: No Ping reply in 180 seconds.]
D_ has joined #ipfs
D_ has quit [Ping timeout: 246 seconds]
D_ has joined #ipfs
MissLavender5032 has quit [Ping timeout: 260 seconds]
MissLavender5032 has joined #ipfs
octect has quit [Ping timeout: 250 seconds]
rardiol_ has joined #ipfs
PerfecT51340 has joined #ipfs
rardiol has quit [Ping timeout: 256 seconds]
negativetwelveDi has joined #ipfs
KeiraT has quit [Ping timeout: 240 seconds]
KeiraT has joined #ipfs
dethos has quit [Quit: WeeChat 2.6]
PerfecT51340 has quit [Ping timeout: 240 seconds]
ipfs-stackbot has quit [Remote host closed the connection]
ipfs-stackbot has joined #ipfs
edouard has joined #ipfs
edouard has quit [Remote host closed the connection]
mauz555 has quit []
edouard has joined #ipfs
edouard has quit [Ping timeout: 264 seconds]
RingtailedFox has joined #ipfs
dqx has quit [Ping timeout: 268 seconds]
dqx has joined #ipfs
fleeky has quit [Read error: Connection reset by peer]
xcm has quit [Remote host closed the connection]
}ls{ has quit [Ping timeout: 256 seconds]
xcm has joined #ipfs
}ls{ has joined #ipfs
stoopkid has quit [Quit: Connection closed for inactivity]
<bonedaddyDiscor4>
> Pinning services are more expensive than cloud storage.
<bonedaddyDiscor4>
Not entirely, https://temporal.cloud is cheaper than digital ocean block storage, and is $0.005 more expensive that google cloud block storage. It's roughly equivalent ot storing all your data in GCE on disks, except its IPFS. https://play2.temporal.cloud
<bonedaddyDiscor4>
> Pinning services are more expensive than cloud storage.
<bonedaddyDiscor4>
Not entirely, https://temporal.cloud is cheaper than digital ocean block storage, and is $0.005 more expensive that google cloud block storage. It's roughly equivalent ot storing all your data in GCE on disks, except its IPFS. https://play2.temporal.cloud for acocunt access
<bonedaddyDiscor4>
> Pinning services are more expensive than cloud storage.
<bonedaddyDiscor4>
Not entirely, https://temporal.cloud is cheaper than digital ocean block storage (50% -> 30% cheaper depending on your account tier) and is $0.005 more expensive that google cloud block storage. It's roughly equivalent ot storing all your data in GCE on disks, except its on IPFS. https://play2.temporal.cloud for acocunt access
<SebastienDiscord>
You gave me one of the most expensive one, but S3 is much cheaper (around $0.023 per GB).
<SebastienDiscord>
You gave me one of the most expensive ones, but S3 is much cheaper (around $0.023 per GB).
<SebastienDiscord>
That's half the best Temporal plan. For very large datasets (those handled by millions users based software), this is simply not an option for now.
fleeky has joined #ipfs
<ShokuninDiscord4>
I mean, it's not like IPFS is a silver bullet
<ShokuninDiscord4>
It's a tech that can work if you want specific decentralization, which is cool
<ShokuninDiscord4>
It has downsides, it's work in progress, but it's definitely not mature like cloud offerings
<ShokuninDiscord4>
I don't think it's a good fit to expect what you'd do on a cloud on ipfs. Rather, I'd look at what can ipfs do and how to leverage that for specific use cases
<bonedaddyDiscor4>
> You gave me one of the most expensive ones, but S3 is much cheaper (around $0.023 per GB).
<bonedaddyDiscor4>
We're talking about costs of IPFS services, of course no IPFS service will be as cheap as cloud storage
<bonedaddyDiscor4>
but $0.05/$0.07 is much better than $0.14 -> $0.30
<bonedaddyDiscor4>
also S3 has bandwidth, and ingress calls too so its not just as simple as $0.023/gb * x gb
<SebastienDiscord>
<span class="d-mention d-user">Shokunin</span> I totally agree and won't go too deep into the discussion. Yet hosting a website is one of the key use cases highlighted in IPFS usage.
<SebastienDiscord>
> also S3 has bandwidth, and ingress calls too so its not just as simple as $0.023/gb * x gb
<SebastienDiscord>
<span class="d-mention d-user">bonedaddy</span> true, but it cannot be evaluated without a proper use case (number of users and requests)
<bonedaddyDiscor4>
that doesn't really make sense tbh, there's a proper use case for everything. If you dont have a use case for whatever it is you're doing, then you're already off on the wrong foot
<SebastienDiscord>
Sorry I meant an exact case, as in an exact client with a number of users and requests.
<bonedaddyDiscor4>
ah ok makes esnse
<bonedaddyDiscor4>
ah ok makes sense
cparker[m] has joined #ipfs
<SebastienDiscord>
I'm not trying to demote the IPFS here, I'm just trying to evaluate what can of business could in the near future migrate from classic cloud storage to this distributed one.
<SebastienDiscord>
I see FileCoin as a potential great solution, but the IPFS alone for now seems unclear to me.
NashtareDiscord[ has joined #ipfs
<bonedaddyDiscor4>
it's valuable in certain cases for a business, like supply chain data storage. But IPFS as a global filesystem for everyone and everything is a dead meme
<bonedaddyDiscor4>
once the various issues are fixed, perhaps it will be a good meme again
<bonedaddyDiscor4>
but until then it's good at a few specific things that are useful to certain types of businesses that dont need to do the whole "global swarm of peers" thing
<swedneck>
oh okay i guess umatrix just got confused
<swedneck>
phew
<eleitl[m]>
why does ipfs add nocopy have no apparent limit on pinning?
<swedneck>
btw what should i do if i have 2+ nodes on a network behind NAT?
<swedneck>
do i just port forward the most reliable?
<eleitl[m]>
Why don't you run a local cluster, and only forward the port to the full node?
<swedneck>
?
<eleitl[m]>
Also, do you have dual-stack? Then just assign public IPv6 addresses, and allow the relevant ports in the firewall.
<swedneck>
i don't have ipv6, sweden way behind on that
<eleitl[m]>
You could use an IPv6 tunnel, but that's not quite the same thing.
<eleitl[m]>
HE.net does free IPv6 tunnels.
<swedneck>
yeah i know, but i'm just wondering how to handle multiple ipfs nodes behind one ipv4 NAT
<eleitl[m]>
You could run one full IPFS node and let the other ones act as a cluster.
<eleitl[m]>
You would only deal with the full IPFS node in the firewall/NAT port forward.
<swedneck>
the problem with a cluster is that it pins on all nodes
<swedneck>
i don't want that
<swedneck>
i want separate nodes
<eleitl[m]>
You wouldn't run a full cluster, but just manual peering with the other nodes. That way you have full control over pinning.
<eleitl[m]>
Basically, a DIY dumb cluster.
<ylp>
interesting, so the front node is basically a cache for other hidden nodes ?
<ylp>
is it possible to configure a node this way ?
<ylp>
or do I misunderstood :)
<eleitl[m]>
It is possible, but I can't give you a config template right away, maybe somebody else can give you a pointer. Or you adapt one of the online examples to this use pattern.
<ylp>
I only have one node but it's something to keep in mind
gimzmoe has quit [Ping timeout: 264 seconds]
vmx has joined #ipfs
<swedneck>
i'm just gonna port forward my server node for now
Adbray has quit [Remote host closed the connection]
Adbray has quit [Remote host closed the connection]
Adbray has joined #ipfs
anyzero has joined #ipfs
anyzero has quit [Client Quit]
anyzero has joined #ipfs
anyzero has quit [Client Quit]
anyzero has joined #ipfs
anyzero has quit [Client Quit]
anyzero has joined #ipfs
anyzero has quit [Client Quit]
anyzero has joined #ipfs
anyzero has quit [Remote host closed the connection]
CGretski has quit [Ping timeout: 246 seconds]
se7en has quit [Ping timeout: 246 seconds]
se7en has joined #ipfs
CGretski has joined #ipfs
<DavidGasquez[m]1>
Hey there! This one might be a dumb question, please bear with me! I'm curious why and how IPFS helps with connectivity. I've read that IPFS makes available things "offline" or with bad connectivity but I was wondering what that does really mean. Initially, I was looking for a way I could browse Internet from a poor connectivity area in the middle of nowhere. I thought about backing up Wikipedia (found Kiwix) but got
<DavidGasquez[m]1>
stuck on how could I browse other places with snapshots (e.g. HN, reddit). I remembered that IPFS could help with that but I can't wrap my mind around it. Does it helps if I have a node? Should I pin my own content from another place and then move that to the bad connectivity place?
nickdex[m] has joined #ipfs
<swedneck>
afaik there's no way to backup ordinary websites with ipfs right now
<swedneck>
most of the "permanence" features really only apply when it's used as the default, which is far from happening yet
rappet has quit [Ping timeout: 246 seconds]
Newami has joined #ipfs
rappet has joined #ipfs
dethos has joined #ipfs
ygrek has joined #ipfs
jcea has joined #ipfs
gimzmoe has joined #ipfs
ipfs-stackbot has quit [Read error: Connection reset by peer]
xcm has quit [Remote host closed the connection]
xcm has joined #ipfs
<DavidGasquez[m]1>
got it! Curious how are using the Wikipedia backup in places where the main website is locked. Is that just through another URL or is more like getting it from the closest IPFS node?
ipfs-stackbot has joined #ipfs
gimzmoe has quit [Read error: No route to host]
gimzmoe has joined #ipfs
Acacia has quit [Ping timeout: 272 seconds]
mowcat has joined #ipfs
Acacia has joined #ipfs
<TraderOne[m]>
Can i list what is in my tiny part of distributed DHT?