pfrazee has quit [Remote host closed the connection]
silotis has joined #ipfs
Guest64299[m] has joined #ipfs
espadrine_ has joined #ipfs
uzyn has joined #ipfs
s_kunk has quit [Ping timeout: 244 seconds]
slothbag has joined #ipfs
<r0kk3rz>
flyingzumwalt: depends on when, but sure
<flyingzumwalt>
r0kk3rz tkorrison[m] check out the github issue. I'm thinking January might work.
<flyingzumwalt>
some UK people are currently working on really cool demos that should be ready to show by then
corvinux has joined #ipfs
mhhf has joined #ipfs
ljhms has quit [Read error: Connection reset by peer]
<Kubuxu>
pjz: Adaptive replacement cache
wallacoloo has quit [Quit: wallacoloo]
sametsisartenep has joined #ipfs
ljhms has joined #ipfs
corvinux has quit [Ping timeout: 250 seconds]
<mhhf>
Hey, I have a simple use case and don't know if ipfs provides the right tools to solve this: I want to push fast and reliably data from peer A to peer B. The data is of course in IPLD format.
<mhhf>
My problem is that A cannot connect to B imediatelly, because B has no capacity to connect to other peers, although A has the data B is looking for. And if they are managed to connect, is there a way to letting A know B has received all data, or should this been done on another protocoll?
<mhhf>
Currently I use simple TCP connections for this.
<mhhf>
Is there some resource I can study to learn more about this?
PseudoNoob has quit [Remote host closed the connection]
soloojos has joined #ipfs
xelra has quit [Ping timeout: 250 seconds]
ygrek_ has joined #ipfs
ygrek has quit [Ping timeout: 250 seconds]
disoxygenation has quit [Ping timeout: 248 seconds]
HastaJun_ has quit [Ping timeout: 260 seconds]
vtomole has quit [Ping timeout: 260 seconds]
cubemonkey has joined #ipfs
Tahsin has joined #ipfs
ygrek_ has quit [Ping timeout: 260 seconds]
<Kubuxu>
<3
soloojos has quit [Remote host closed the connection]
soloojos has joined #ipfs
lkcl has joined #ipfs
<dignifiedquire>
lgierth: whyrusleeping haad is there a way to subscribe to the same topic through two different clients?
<dignifiedquire>
I'm currently getting "error, duplicate subscription" if I connect to the same node using two different browsers
<dignifiedquire>
when doing a pubsub subscription
anonymuse has joined #ipfs
poga has quit [Ping timeout: 268 seconds]
anonymuse has quit [Remote host closed the connection]
poga has joined #ipfs
<A124>
How large are the caches on the gateways?
<A124>
We got onto heated discussion today, where I was even asked to prove that if we do not host the content it will cease to exist.
<WinterFox[m]>
What advantages does ipfs have over torrents?
<A124>
As per IPFS spec or currently?
<A124>
Maybe I should remain silent, there are too many to name.
<WinterFox[m]>
The spec I guess
<A124>
I could only recommend reading it then.
<Kubuxu>
A124: about 20Gig or something like that.
<WinterFox[m]>
I only know of some little ones like deduplication everywhere
<daviddias>
A124 it depends
<Kubuxu>
That isn't much.
<daviddias>
We have some beefy gateways too, but rule of thumb is that you should not use a gateway as your cache
<A124>
No, it is not, but [he] put video to his node, loaded via gateway, shutdown his node and it worked.
<Kubuxu>
We have beefy storage servers, gateways are quite smaller.
<Kubuxu>
Yeah, because it would immensely more expensive to not store it and just fetch it every time.
<ansuz>
> 09:23 < A124> Maybe I should remain silent, there are too many to name.
<Kubuxu>
but it might not be a bad idea after all.
<ansuz>
where have I seen that before...
PseudoNoob has joined #ipfs
<Kubuxu>
Hmm, maybe even create custom GC that removes content from gateways explicitly after 24h
<Kubuxu>
of not being used
<A124>
Kubuxu how do you... decide what is on storage servers? Only what you set it to, or another magic?
<A124>
ansuz Did I say something like that in past, or are you making a reference or just wondering?
<Kubuxu>
what we explicitly put on them or we sometimes give access to community members to run archival projects on them.
<A124>
Can I get such access?
<A124>
Or... is it even managed who pins what?
<A124>
The GC policy is not a bad idea at all. Disabled by default.
<A124>
And with a tunable time window.
<Kubuxu>
Yeah, it would be a good thing for public gateways.
<A124>
How does the current GC pick what to evict on overfill?
<Kubuxu>
everything that is not pinned
<A124>
Even the auto-gc?
<Kubuxu>
yus
<Kubuxu>
yes
<A124>
Ouch.
<Kubuxu>
we use mark-and-sweep right now
<Kubuxu>
it isn't the best but it works, there are plans for future but beyond our current bandwidth
<A124>
One thing that might be used/abused is just daily refreshing some assets.
<A124>
So that remain in cache.
<Kubuxu>
They will be downloaded to the cache when someone uses them
<A124>
What is the usual gateway (aggregated?) B/W?
<Kubuxu>
also hashes pinned by pinbot in this channel
<A124>
Active B/W, not available.
<Kubuxu>
are being pinned on GW
<haad>
dignifiedquire: unfortunately not :/ the subscriptions are tracked in js-ipfs-api on a global level...
<dignifiedquire>
okay :/
<Kubuxu>
the http traffic or the ipfs traffic?
<ansuz>
A124: I was looking for a jpeg of it, but I couldn't find it
<dignifiedquire>
that makes testing really nasty
<haad>
yup :/
<ansuz>
A124: somebody recommended Rewriting It In Rust, and they said they wouldn't even go over the benefits because they were obvious
<ansuz>
not sure what project
HastaJun_ has joined #ipfs
<A124>
Oh.
Boomerang has joined #ipfs
<ansuz>
yea
<ansuz>
lol
<A124>
Kubuxu Well, I got different things I would like to pin and unpin, mostly devel stuff, html, and such. ... Which brings me one more question... if multiple things reference same blocks or other things, when one does pin rm -r, are they removed also?
<A124>
Or is there a reference counter?
<Kubuxu>
we keep only the roots of the pins
<Kubuxu>
then we traverse the whole file for block refs on GC event
<A124>
ansuz Read your article, informative, educational and pleasant to read. RIIR!
<Kubuxu>
this creates list of blocks to keep
<A124>
Kubuxu I mean pins, not GC in this case.
<Kubuxu>
as I said: we keep only roots of the pins
<Kubuxu>
so if you pin A and you pin B and they share C
matrixmatrixorg[ has left #ipfs ["User left"]
<Kubuxu>
GC when run will look at your pin list ([A, B]) and traverse it collect all blocks that should be kept [..., C, ...]
<Kubuxu>
this means that it doesn't matter if you unpin A, C will be kept
<dignifiedquire>
daviddias: haad having problems with cors :8
<dignifiedquire>
:Ö(
<ansuz>
A124: I got shit from cjd and lgierth about it not being controversial enough
<Kubuxu>
yeas, it will be Pre-shared symatric key only
<Kubuxu>
for now
<Kubuxu>
why are you asking?
<kevina>
I just could not tell from the description.
<kevina>
As I am unfamiliar with the terminology used.
<kevina>
For example ECDH and KeyStretcher
<Kubuxu>
pre-shared key in 99% means symmetric encryption
<kevina>
okay, thanks for the clarification
<Kubuxu>
so ECDH is elliptic curves diffie-hellman, which allows to generate shared-secret from priv and pub key pair
<Kubuxu>
and then after you have one shared secret you would use symmetric encryption after that.
<kevina>
I see, now it makes sense.
<kevina>
so I take it that this is an all or nothing for now
<kevina>
i.e. an ipfs node will not talk to both public and private nodes at the same time
ligi has quit [Ping timeout: 252 seconds]
<Kubuxu>
no, you are either part of one priv network or you are part of public netwrok
<kevina>
okay that what I thought, sounds good
<ansuz>
you're either with us or against us
<kevina>
one of these days I really need to learn more
<Kubuxu>
I am in process of doing a writeup on crypto that would satisfy our demands and would have minimal overhead
<kevina>
I know of the basic concepts of how public-key infrastructure and related works, but none of the specifics
ligi has joined #ipfs
<dignifiedquire>
flyingzumwalt: pretty sure I got things figured out on the plane
alu has quit [Quit: WeeChat 1.0.1]
<dignifiedquire>
so I will be able to push a working version tomorrow when I am back on the internet for you to demo
<dignifiedquire>
(hopefully)
flyingzumwalt has quit [Quit: Connection closed for inactivity]
<Kubuxu>
whyrusleeping: but either way, I can proceed without transport layes
<Kubuxu>
just do it the same way I wanted to do with secio
<Kubuxu>
as I think storing secret inside the multiaddr that is exposed everywhere across the API and network isn't the best idea
bastianilso__ has joined #ipfs
d10r has quit [Ping timeout: 250 seconds]
ligi has quit [Ping timeout: 245 seconds]
azdle has joined #ipfs
Mizzu has quit [Quit: WeeChat 1.5]
kode54 has quit [Quit: ZNC 1.6.3+deb1 - http://znc.in]
chris613 has joined #ipfs
rendar has quit [Quit: Leaving]
espadrine has joined #ipfs
maxlath has quit [Quit: maxlath]
ashay has joined #ipfs
<ashay>
Hey @haad Wanted to play around with orbit-db for a project and was having some trouble getting the examples working
kode54 has joined #ipfs
maxlath has joined #ipfs
d10r has joined #ipfs
ashay_ has joined #ipfs
ashay_ has quit [Client Quit]
ligi has joined #ipfs
mildred1 has quit [Quit: Leaving.]
patcon has quit [Ping timeout: 252 seconds]
Dao has joined #ipfs
<dignifiedquire>
hey ashay haad is currently travelling as far as I know, but if you could post your issues in a gh issue we can take a look at what needs improving
Boomerang has quit [Quit: Leaving]
taaem has quit [Read error: Connection reset by peer]
taaem has joined #ipfs
gavinmcdermott has joined #ipfs
<Kubuxu>
whyrusleeping: should I just continue one the salsa20 pre-secio, without transport-layers?
<whyrusleeping>
Kubuxu: ACK
<whyrusleeping>
i'd go ahead and start on that
<whyrusleeping>
but try and get some review
<whyrusleeping>
maybe shame cjd into looking at it
<ansuz>
lol
<Kubuxu>
yeah, I will try, this crypto is almost the same as cjdns is using
<ansuz>
glhf
<Kubuxu>
cjd hates reading crypto IIRC
<ansuz>
almost as much as he hates ancaps
<whyrusleeping>
Kubuxu: thats why i said "shame him into doing it"
<ansuz>
:D
DiCE1904 has quit [Remote host closed the connection]
polezaivsani has quit [Remote host closed the connection]
bastianilso__ has quit [Quit: bastianilso__]
anonymuse has quit [Remote host closed the connection]
anonymuse has joined #ipfs
_fil_ has joined #ipfs
matoro has joined #ipfs
anonymuse has quit [Ping timeout: 250 seconds]
Dao has quit [Quit: Page closed]
taaem has quit [Ping timeout: 265 seconds]
espadrine has quit [Ping timeout: 260 seconds]
IlanGodik has quit [Quit: Connection closed for inactivity]
jedahan_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<kumavis>
can someone explain why "The unified multicodecs theory" is a good idea
cemerick has quit [Ping timeout: 244 seconds]
<ansuz>
git uses SHA1, and can't easily upgrade
<ansuz>
s/easily//
<kumavis>
seems that putting different types of things (binary encoding type, hash fn) into the same limited space, we are going to run out of that limited space
<gavinmcdermott>
question about IPNS and name resolves: I name published some content under a peerId about a month or so ago and today, when 'name resolving' that id, it couldn't resolve it.
* ansuz
sleeps
<gavinmcdermott>
I ended up needing to republish under that id
<Kubuxu>
gavinmcdermott: the records by default have validity of 24h, there are many reasons for that
<gavinmcdermott>
perfect
<gavinmcdermott>
that's what I was looking for
<gavinmcdermott>
gracias @kubuxu
<Kubuxu>
Awesome that I could help.
<Kubuxu>
kumavis: he is unhappy because the eth will have to use the varint
<gavinmcdermott>
if you know off hand - where in the larger ipfs repo ecosystem can I look into this...if not, I'll find it
<kumavis>
if space was unlimited, this would not be an issue
<kumavis>
or is this a matter of high value (short name) vs low value space (long name)
<Kubuxu>
the latter
<Kubuxu>
the about 0x80 will be a byte longer
<Kubuxu>
as for the varint
<Kubuxu>
s/about/above
palkeo has quit [Quit: Konversation terminated!]
<Kubuxu>
gavinmcdermott: I think the time to live of a record can be set in the config
<Kubuxu>
but also nodes will keep records only for limited time too
<Kubuxu>
IIRC
<gavinmcdermott>
Also: if a node is resolving on IPNS for its own id, does it hit the network for that or pull it out of some local cache / repo?
<Kubuxu>
it hits the network
<gavinmcdermott>
thanks again!
jedahan has joined #ipfs
<kumavis>
@Kubuxu if they are varints, then they would only be >1 byte if they are >2**7 (128), correct?
<kumavis>
er, (2**7)-1
<Kubuxu>
yeah, so everything >=0x80 is to bytes
<Kubuxu>
s/to/two
<kumavis>
right
<kumavis>
ok
<Kubuxu>
gavinmcdermott: IPNS puts very high priority on the C in CAP theorem, this is something that is not tunable right now
<Kubuxu>
it might be when we redesign IPNS as IPRecordSystem
<kumavis>
Kubuxu: so i still dont understand why unrelated things (binary encoding type, hash fn) are sharing the same namespace
<Kubuxu>
because we use the same encoding in many places
<Kubuxu>
<ID><Length><data>
<Kubuxu>
if we make the ID namespace shared
<Kubuxu>
then it is more descriptive
<Kubuxu>
as you don't have to know which namespace is being used
<Kubuxu>
I am off for today, gn.
<kumavis>
Kubuxu: thanks
ashay has quit [Ping timeout: 260 seconds]
ygrek has joined #ipfs
neurrowcat has quit [Quit: Deebidappidoodah!]
kobigurk has quit [Ping timeout: 252 seconds]
kulelu88 has quit [Ping timeout: 256 seconds]
captain_morgan has quit [Read error: Connection reset by peer]