<whyrusleeping>
ylp1: cjdns also makes you anonymous
<cryptix>
does it?
<cryptix>
pseudonymous at best
<cjd>
are you sure you're not thinking of madesafe?
<whyrusleeping>
lol
<whyrusleeping>
right
<cjd>
What I never understood was how maidsafe is able to defeat evil maid attacks which use cold boot
<cryptix>
also the routing should be easy to see
JasonWoof has joined #ipfs
<ralphtheninja>
daviddias: is the js version of ipfs "usable"?
<ralphtheninja>
cjd: it could also be dormir I guess
<daviddias>
ralphtheninja: usable as in "it does somethings", yes
<cjd>
yeah that too
<Guest51759>
does anyone wanna play warframe?
<whyrusleeping>
ralphtheninja: usable, yes. but for what purpose we don't know
<daviddias>
ralphtheninja: but to develop apps on top of IPFS, use go-ipfs and js-ipfs-api for example
<daviddias>
it packs all the features
<ralphtheninja>
daviddias: roger
Guest51759 is now known as Vegemite
Vegemite is now known as Vegemite_
<ralphtheninja>
daviddias: but I want to build on top of js-ipfs :D
<daviddias>
help us build js-ipfs :D
<ralphtheninja>
I'd love to
<ralphtheninja>
daviddias: oh you really want it to run in the browser and not just in node
<ralphtheninja>
nice
<daviddias>
the goal is to be able to do both
<ralphtheninja>
nodnod
<daviddias>
but yes, the reason why it started was to make it run on the browser :)
<cjd>
WebRTC
<whyrusleeping>
webarrteesee
<cjd>
looks like a state in the sawouth
<cryptix>
re webrtc: does windows have native sctp?
<Vegemite_>
ralphtheninja: i sent you a private message
cryptix has quit [Quit: leaving]
Senji has quit [Disconnected by services]
Senj has joined #ipfs
Senj is now known as Senji
cryptix has joined #ipfs
s_kunk_ is now known as s_kunk
s_kunk has quit [Changing host]
s_kunk has joined #ipfs
<whyrusleeping>
lgierth: can biham be set up so that the daemon daviddias is using is different than the one anyone else would be using?
Shibe has quit [Remote host closed the connection]
multivac has quit [Remote host closed the connection]
<ralphtheninja>
what's biham?
m0ns00n has joined #ipfs
<whyrusleeping>
its a storage node
yunconven has quit [K-Lined]
jhulten has joined #ipfs
rombou has quit [Ping timeout: 264 seconds]
jhulten has quit [Ping timeout: 246 seconds]
cemerick has joined #ipfs
Guest76511 has joined #ipfs
jaboja has quit [Ping timeout: 240 seconds]
<daviddias>
cryptix: WebRTC uses SCTP on top of UDP
jaboja has joined #ipfs
hellertime has joined #ipfs
hoony has quit [Remote host closed the connection]
Vegemite has joined #ipfs
<ipfsbot>
[go-ipfs] whyrusleeping created cleanup/mfs-dir (+1 new commit): http://git.io/vzejJ
_whitelogger has joined #ipfs
<cjd>
If you want to not be blocked by every nat/firewall on earth, you need to use TCP or UDP
<cryptix>
okay, i knew that about "pure" sctp - but they layer webrtc in another layer of normal/legacy ip/udp?
<cryptix>
(i guess so that you dont need a native sctp stack in the OS?)
<The_8472>
indeed
cemerick has quit [Ping timeout: 276 seconds]
<cryptix>
okay, i see - "awesome"
<cjd>
no, it needs to be under UDP because otherwise everything would block you
<cjd>
but "under UDP" just means you pay 8 bytes per packet, it's really not a big deal
<Kubuxu>
cjd: There is nothing bad with using TCP/UDP and SCTP. As adaptation of SCTP rises, more hosts will peer with SCTP
<cjd>
SCTP over TCP would be bad
<cjd>
SCTP over UDP is no different from SCTP without UDP, just costs +8 bytes
<cryptix>
Kubuxu: i made that point out of purist stubbornness - i'm sure it works somehow (see goog hangover.. :)
<Kubuxu>
Not SCTP over TPC but when SCTP is available bind it also
<The_8472>
think of udp ports as user-mode protocol numbers as opposed to ip having kernel mode protocol numbers :P
<cjd>
and udp has 65k of them and the IP header only allows 256
<Kubuxu>
As well as pure TCP and maybe UDP/µTP
<cjd>
but everyone wants their protocol to have an "official designation" with it's own IPv4->nextHeader number, even though in practice it makes it useless because firewalls and nats passthrough TCP and UDP
<cjd>
IPSec anyone ?
<Kubuxu>
Yes but SCTP is already established protocol. They are not crafting anything new.
<The_8472>
fsvo passthrough. ipfs still fries my router. haggling with my ISP about getting bridge mode to work (again)
<The_8472>
and of course CGN
<Kubuxu>
µTP is fairly new and uses UDP as transport layer.
<cjd>
so is IPSec, in either case try telling that to your DSL modem :)
<Kubuxu>
but if you have hosts with SCTP capabilities, why not connect them over SCTP, just give the option.
<whyrusleeping>
The_8472: have you tried using utp on 0.4.0?
<cjd>
Why should SCTP be in kernelspace?
<The_8472>
whyrusleeping, not yet, but I intend to.
<cjd>
or TCP for that matter
<achin>
any of you running v0.4.0? if so, can you try `ipfs refs QmSyKjMk3N2NCZrDYUMS3EwBENsgV39G5zGXgkCDvfZW8W` ?
<ipfsbot>
[js-ipfs] diasdavid pushed 2 new commits to feat/bootstrap: http://git.io/vzvJl
<ipfsbot>
js-ipfs/feat/bootstrap ef701dc David Dias: add bootstrap list to cli
<ipfsbot>
js-ipfs/feat/bootstrap 467ef34 David Dias: bootstrap add rm list complete + cli + cli tests
<Kubuxu>
It already is since 2.6
<The_8472>
sendfile?
<Kubuxu>
I can bind UTP but IPv6 only.
<The_8472>
although i guess scatter-gather IO with mmaped files gives you most of the benefits of sendfile
_whitelogger has quit [Ping timeout: 240 seconds]
_whitelogger_ has joined #ipfs
<cjd>
libtcp.so ?
<cjd>
nobody who isn't really on top of things will even get the handshakes right for tcp, never mind the congestion control
reit has quit [Ping timeout: 240 seconds]
<The_8472>
libtcp.6.1.so or libtcp.7.3.fixnagle.so?
<cjd>
pick one.. if you rely on the kernel you'll have both
<cjd>
and windows_weird_tcp.dll
<cjd>
Or you rely on nodejs and golang to come packaged with mean nasty ones by default
<The_8472>
well, one issue is how do you manage flows?
<The_8472>
i.e. how do you deliver things to the right process?
<daviddias>
ralphtheninja: couldn't understand what "d:o" means??
<cjd>
As I said, I think the UDP "port number" should have been in the IP header
<daviddias>
Agreed
<daviddias>
what you mean by "mean nasty ones" on Go and Node?
<cjd>
fast
<The_8472>
what about other IP things owned by the kernel? network management. routing, neighbor discovery etc.
<The_8472>
they're done via ICMP, which also is an IP protocol number
<The_8472>
you'd have to do a lot of redesigning if you want userspace to manage it. privileged port ranges i guess?
<cjd>
In general I'd say if it'd a cold codepath it ought to be out in userspace... but that's a bit microkerneley
<ipfsbot>
[js-ipfs] diasdavid deleted feat/bootstrap at 467ef34: http://git.io/vzvIg
<Ape>
Every application could have its own IPv6 address
<The_8472>
i'm doing that with firejail
<cjd>
That would help, as long as you have a big enough block of addresses
<The_8472>
spawns a new network namespace with v6 address.
<Ape>
What is mfs?
<cjd>
Other approach is to shove everything into kernelspace but for that we need some kind of safe ASM which cannot touch the wrong pages
<cjd>
same result though
<Kubuxu>
Ape: it is a way to build directory structure and files inside of IPFS, like in fuse but without fuse.
<Ape>
How can I use it then?
<Kubuxu>
ipfs foles
<ansuz>
foles api
<ansuz>
new in 0409
<ansuz>
040*
<whyrusleeping>
lol
<Kubuxu>
ipfs files
<whyrusleeping>
fips flies
<Kubuxu>
Stupid autocorrect in phone.
<Ape>
ansuz has the same one :D
<ralphtheninja>
daviddias: "d:o" is an abbreviation of ditto
<daviddias>
got it, I had a typo! :)
rombou has quit [Remote host closed the connection]
<Kubuxu>
It is really awesome
<ralphtheninja>
daviddias: two typos :)
<ralphtheninja>
back to the couch .. still sick :/
<ansuz>
will ipfs correct my typos?
<ansuz>
possibly at the same time it's making me anonymous
<whyrusleeping>
ansuz: only if you log into facebook and click 'make me anonymouse'
<ansuz>
that's pretty easy
<ansuz>
nice work
<cjd>
IPFS Never Forgives, IPFS Never Forgets
<ansuz>
I'll just need to make a facebook account
<Kubuxu>
Because before you couldn't easily build in IPFS structures. My plan is to use it for deploying my blog and keep history alive.
<The_8472>
cjd, in the end the differences and benefits seem minor. what I would kill for would be publicly routed multicast. but that seems to be a hard problem
fiber has joined #ipfs
<cjd>
Multicast and TCP go hand-in-hand
<The_8472>
heh
<cjd>
because guaranteed delivery
<The_8472>
udp obviously
<cjd>
no I mean multi over TCP
<The_8472>
what?
<cjd>
like IRC or whatever kind of websocket/feeds
<The_8472>
no, that's poor man's multicast
rombou has joined #ipfs
<The_8472>
it does not give you any of the deduplicating benefits
rombou has quit [Client Quit]
<cjd>
but you notice these poor men never use fire-and-forget for their multicast :)
<The_8472>
or rather: rich man's multicast
<The_8472>
retrieving the bits and pieces that got dropped en route can still be done via unicast. the point of using multicast is having efficient delivery
<The_8472>
youtube from your basement so to speak
<cjd>
hm maybe
<Kubuxu>
whyrusleeping: it would be great to have in files CLI stat that gives you only the hash. Either as separate command or as a flag.
<cjd>
identifying the people who want to access resource X is easy, identifying the people who need to have datagram Y is hard
<Kubuxu>
IDK if you have seen it on GH.
<cjd>
*need to have datagram Y *right now*
<The_8472>
yeah, lots of state and probably some fuzzy approaches needed. allow a little over-delivery to the next hop and hope they have the necessary state to prune it away or something like that
<whyrusleeping>
Kubuxu: yeah, the files api is still fairly under construction at the momeny
<whyrusleeping>
moment*
fiber has quit [Ping timeout: 240 seconds]
<whyrusleeping>
its got some fun stuff i need to work out so that garbage collection doesnt murder it
jgraef2 has joined #ipfs
<Kubuxu>
Everything under MFS will be pinned I suppose?
<The_8472>
i wonder what hypothetical hardware would be needed to support current SSM routing protocols at the internet core
jgraef has quit [Ping timeout: 240 seconds]
<Kubuxu>
whyrusleeping: if I wanted to implement resolving of _dnslink. should I PR it into v04 or master?
jgraef has joined #ipfs
jgraef2 has quit [Ping timeout: 265 seconds]
Encrypt has quit [Quit: Quitte]
<whyrusleeping>
Kubuxu: resolving of dnslink for what?
<whyrusleeping>
all new PR's should go into 0.4.0
<whyrusleeping>
Kubuxu: yeah, everything under mfs will be pinned, but for performance, we have a '--flush=false' flag you can pass that skips the sync to disk
<whyrusleeping>
to avoid tons of needless hashing
<Kubuxu>
There was proposal to make "_dnslink.example.com" resolve in case of resolving example.com
<ipfsbot>
[js-ipfs] diasdavid pushed 1 new commit to master: http://git.io/vzv8C
<ipfsbot>
js-ipfs/master b8edf6e David Dias: fix typo
m0ns00n has quit [Quit: undefined]
<Kubuxu>
It is useful if you want to CNAM main record (which you shouldn't do in case of SLD but you can in case of third level domain.
<whyrusleeping>
hrm, okay
<Kubuxu>
I can find the issue on /notes when I get on my notebook.
elima has quit [Ping timeout: 272 seconds]
nicolagreco_ has joined #ipfs
jgraef has quit [Quit: Bye]
jgraef has joined #ipfs
boltmaker has joined #ipfs
<cryptix>
lgierth: can you show me the domain rewrite trick? ie serving 32c3.ipfs.io from dnslink=/ipfs/.. entry
<cryptix>
lgierth: also the fahrplan at 32c3 is out of date (1.5a is latest, i hope its the last... :)
<cryptix>
at 32c3.ipfs.io *
<lgierth>
whyrusleeping: yeah we can do that (ipfs per user), just makes it a bit more complicated. is it too hairy to run everything in one instance?
<lgierth>
cryptix: haha actually i removed 32c3.ipfs.io just the other day
<cryptix>
okay :)
<lgierth>
so, you set a TXT record "dnslink=<ipfs-path>", and one or more A records pointing to gateways
<lgierth>
however without the A records, you can still resolve /ipns/eaxmple.com
<cryptix>
oh i thought there was another trick involved.. didn't know the http gw can do it itself
<lgierth>
ipns has pluggable resolution mechanisms -- icann dns is the first
<lgierth>
we also want .bit for example
<cryptix>
yea, i know - i just thought *.ipfs.io needed some nginx magic in front
<cryptix>
okay awesome
nicolagreco_ has left #ipfs [#ipfs]
<achin>
lgierth: if you have a few moments today, i'd love to help in figuring out why a directory won't load on v04x.ipfs.io. i'm suspecting it has to do with how large it is
<lgierth>
the only nginx magic is to make /ipfs /ipns and /api work on ipfs.io, although it is *also* a dnslink page
<whyrusleeping>
lgierth: the way daviddias is using it, if the daemon is rebooted before he calls 'ipfs files flush' then he loses data
<lgierth>
and /refs
<whyrusleeping>
so he wants to minimize the risk of that happening
<lgierth>
ouch yeah
elima has joined #ipfs
<lgierth>
achin: i'm not quite sure how to debug it, but i can scan the logs a bit with you
<lgierth>
i'll be at the office in ~1h
<achin>
super, thanks
* lgierth
-> supermarket
<lgierth>
M-davidar: yay replication \o/
<ipfsbot>
[go-ipfs] whyrusleeping pushed 1 new commit to feat/mfs-flush-cmd: http://git.io/vzvwI
<Kubuxu>
lgierth: Yeah, with goroutines, I hope that net.lookupTXT goes to sleep when waits for DNS response?
<Kubuxu>
yeah it does.
Soft has quit [Quit: WeeChat 1.4-rc1]
<Kubuxu>
It might won't work on Windows though.
Encrypt has joined #ipfs
<Kubuxu>
As on Windows it resolves using Native library.
<Kubuxu>
lgierth: which should I resolve first then? _dnslink.example.com or example.com?
<lgierth>
_dnslink
<lgierth>
and if that doesn't exist fall back to example.com
Soft has joined #ipfs
<Kubuxu>
I will use goroutines either way just in case of non-Unix it will work as sequential.
jgraef has left #ipfs ["Leaving"]
jgraef has joined #ipfs
jholden has joined #ipfs
<jgraef>
Does anyone know how I use "ipfs object patch *" over HTTP?
<jgraef>
What counts as path component and what as argument?
neurosis12 has joined #ipfs
elima has quit [Ping timeout: 250 seconds]
<jgraef>
Ah okay. I see what's going on. For add/rm-link it works with the object key being part of the path and the sub-command and the rest being arguments. For set-data the API complains that "args" is missing.
<jgraef>
But ipfs object set/append-data has not args
<jgraef>
ipfs object patch set-data takes an argument - the data. But the data should actually be taken from the request body
Matoro has quit [Read error: Connection reset by peer]
hoony has quit [Remote host closed the connection]
<ipfsbot>
[go-ipfs] Kubuxu opened pull request #2184: Make dns resolve paths under _dnslink.example.com (dev0.4.0...feature/dnslink) http://git.io/vzfGs
<Kubuxu>
lgierth: ^^
<Kubuxu>
achin: It is awesome.
Tv` has joined #ipfs
flounders has joined #ipfs
<jbenet>
yo cryptix
<jbenet>
havent seen you in ages! :) -- did you hang out with people at 32c3?
<ralphtheninja>
achin: why isn't it on ipfs? :)
<jbenet>
davidar around?
<ralphtheninja>
jbenet: we missed you at 32c3
<ralphtheninja>
you need to come next time!
<jbenet>
ralphtheninja: thanks-- indeed i do! i missed you all too :(
jhulten has joined #ipfs
cryptix has quit [Ping timeout: 250 seconds]
cryptix has joined #ipfs
jhulten has quit [Ping timeout: 240 seconds]
<dignifiedquire>
whyrusleeping: fuuuck I got the WNDR4300v2 and that has no support for openwrt -.-
<ralphtheninja>
ccc is interesting in many ways, on one hand it's about hacking and tech etc and on the other hand it feels conservative
<lgierth>
dignifiedquire: yeah? one of the new devices where tp-link blocks flashing alternative firmware?
<dignifiedquire>
lgierth: netgear, no mention of blocking but the wiki and forums say nobody tried, even though it's nearly 2 years old :(
<lgierth>
ah wNdr
<dignifiedquire>
lgierth: any ideas if I should just try to flash it? or if I should just send it back :/
<ralphtheninja>
dignifiedquire: which wndr?
<jbenet>
i thought N was the same as something else-- or maybe it was D..
<dignifiedquire>
ralphtheninja: wndr 4300 v2
<ralphtheninja>
dignifiedquire: that's what I have, maybe not v2 though
ulrichard has quit [Quit: Ex-Chat]
<ralphtheninja>
dignifiedquire: you might need to flash an older version of openwrt and once you have done that you can upgrade
<dignifiedquire>
ralphtheninja: not sure I understand on the wiki they say that v2 is a completely different machine than v1, they just reused the name
<jgraef>
mec-is I've implemented a reader/writer for protobuf in Python3 myself.
mec-is has quit [Ping timeout: 252 seconds]
<drathir>
feross: webrtc atm is the good source of leakage personal data...
<feross>
drathir: not really. those concerns are overhyped
<lgierth>
ok i'm here
<lgierth>
posting
<feross>
drathir: adding webrtc to serviceworker won't change the situation there anyway
<dignifiedquire>
here
m0ns00n has quit [Quit: undefined]
The_8472 has quit [Ping timeout: 240 seconds]
cryptix has quit [Ping timeout: 250 seconds]
<richardlitt>
awesome
<drathir>
feross: yea im sure there are advantages over webrtc for sure, but nice to also possible dynamically manage usage of this function...
<Kubuxu>
I agree with drathir: I want my browser allow me to choose if I want to send.
<feross>
drathir: with the recent browser changes, nothing is leaked
<Kubuxu>
I don't have nothing against it when I am at home or in office, but when I am on mobile connection ....
<feross>
drathir: thats a completely different argument than privacy
<feross>
^sorry meant that for Kubuxu
<daviddias>
here now
<feross>
if a site wants to, it can already make an XHR request and download/upload tons of garbage data from a page
<drathir>
btw about support of tor the other services like i2p or cjdns consider also as native supported connection?
<lgierth>
drathir: yes, it's both work in progress
<tperson>
Hey dignifiedquire, do you know of a http request module that starts with an 'ax' (I know super descriptive). I feel like ive seen it mentioned a lot as of recent as a nice promised based request lib.
<richardlitt>
OK
<Kubuxu>
feross: Yes it can but it is no use for them. The moment it starts being useful they would start using it.
<richardlitt>
We are going to start the sprint now, sorry for the incoming flood of texts
<dignifiedquire>
tperson: that's a great description :D
<richardlitt>
All in favor say aye.
<dignifiedquire>
aye
<lgierth>
yeah aye
<feross>
Kubuxu: you should install a browser extension and just block it if it bothers you
<daviddias>
richardlitt: aye (writting my checkin now)
<dignifiedquire>
tperson: but I'M afraid I don't know any library with "ax" inside, except for "ajax" ;)
* whyrusleeping
is here
<richardlitt>
In order of appearance, I will go first.
<richardlitt>
## This Week
<richardlitt>
Summary: Not a very productive week for me this week, for a variety of reasons, mostly travel related and because it was a short week and I had a lot of admin stuff. I know what I'm doing, just running behind a bit.
<richardlitt>
- [x] Sprint Management
<richardlitt>
- [~] Write the weekly roundup; launch it.
<richardlitt>
- [ ] IPFS API
<richardlitt>
- [x] Split into current functions `master` branch and `expected` spec branch.
The_8472 has joined #ipfs
<richardlitt>
- [ ] All of the rest, see the issue for details
<dignifiedquire>
- [x] Lots of PRs from @RichardLitt that I don't have the links to anymore
<dignifiedquire>
EOF
<richardlitt>
Good work, dig!
<richardlitt>
Thanks so much for the PR review
oszi has joined #ipfs
<richardlitt>
Feels good to have that stuff closed
<dignifiedquire>
richardlitt: I see PR, I review :P
<dignifiedquire>
(at least I try)
<ralphtheninja>
what does a ~ mean?
<dignifiedquire>
partially done
<ralphtheninja>
k
<dignifiedquire>
If anybody has experience with wikipedia xml dumps please ping me I got questions
<richardlitt>
ralphtheninja: sorry, kind of a confusing process. :)
<lgierth>
dignifiedquire: yay do something with all that free disk space on biham ;)
<ralphtheninja>
richardlitt: no worries, I'll just ask if there's something I don't understand :)
<dignifiedquire>
lgierth: when I'm done there will be no space left :P
<lgierth>
good boy
<lgierth>
:)
<jbenet>
difnifiedquire: quick feedback on site-- can you make on a:hover underline? can barely see the glow.
<dignifiedquire>
jbenet: not sure where/what you mean
<jbenet>
(well not a:hover since the left is also a: i think-- just the links in the main section i guess?)
<dignifiedquire>
you mean distributions?
<jbenet>
i cannot tell "Docs, Changelog, All versions" is a link
<jbenet>
yeah
<dignifiedquire>
that's why they are blue ;)
<jbenet>
(underline makes it much more clear + active)
<dignifiedquire>
but okay will take a look
<jbenet>
not all blue text ever is a link.
<jbenet>
dignifiedquire: good stuff with stackoverflow!
<dignifiedquire>
jbenet: thanks, the really cool stuff will come when can mirror wikipedia..but that's a little longer way
<jbenet>
yeah
<richardlitt>
:)
ygrek has joined #ipfs
<dignifiedquire>
jbenet: have you talked to anyone related to wikipedia in the past? cause I need to contact someone about how to do the mirroring
<dignifiedquire>
(their wiki pages are somewhat outdated and say to send them an email)
<richardlitt>
dignifiedquire: not sure. We can figure that out later.
<richardlitt>
for now, thanks. lgierth: you ready?
<lgierth>
yeah, c&p inc
<lgierth>
I've been away for a day visiting my grandma (she turned 80). The ipfs.io v03x/v04x thing is on its way, and I'm very happy to start getting rid of Ansible.
<lgierth>
- 0.4.0
<lgierth>
- [x] v04x.ipfs.io and v03x.ipfs.io
<lgierth>
- [x] more v03x nodes (3 out of 8 now) ipfs/ops-requests#3
<lgierth>
- [ ] discuss migration plan with users
<lgierth>
- [ ] have ipfs.io resolve both v03x and v04x (whichever responds success fast)
<lgierth>
- [x] dead-simple replacement for ansible (aka Provsn)
<lgierth>
- [x] project-repos.ipfs.io dns ipfs/ops-requests#9
<lgierth>
- [x] work through email backlog
<lgierth>
- wishlist
<lgierth>
- ~~set up buildbot worker for Tahoe-LAFS~~ (didn't hear back)
<lgierth>
didn't make as much progress with v03x/v04x as i'd liked, but meh
<Ape>
Should that be "more v04x nodes"?
<lgierth>
eeeh yeah!
<lgierth>
good catch
<richardlitt>
Hooray for the email backlog.
<ralphtheninja>
lgierth: how many nodes are there in total?
<whyrusleeping>
ralphtheninja: nodes in our infrastructure? or in the entire network?
<dignifiedquire>
lgierth: what is provsn Oo
<ralphtheninja>
whyrusleeping: our infrastructure
<ralphtheninja>
whyrusleeping: noticed I said "our" as well :D
<lgierth>
13 nodes in total, plus 2 nodes which i brought to 32c3 which are now waiting to be deployed within freifunk
<jbenet>
dignifiedquire: yes i have-- i met some people they told me to go to their IRC. they seemed interested but conservative/slow. i suggested we start with wikimedia commons (like all the images + video, etc) where there's the biggest gain for them. and which we need before the text looks good.
<whyrusleeping>
ralphtheninja: i think just eight bootstrappers, and a few other storage nodes
<whyrusleeping>
ralphtheninja: lol
<lgierth>
dignifiedquire: provsn is infrastructure provisioning in less than 100 lines of shell
<jbenet>
lgierth: congrats on grandma 80!
<lgierth>
i'll put it in a proper repo this week
<lgierth>
haha thanks i'll pass that on jbenet
<dignifiedquire>
lgierth: that explains why google didn't know it :D
<dignifiedquire>
lgierth: you should make it 80, in honour of your grandma
<jbenet>
lgierth: re nodes, can we have all bootstrappers on both networks? (i.e. 2 independent ipfs nodes with same key)
<whyrusleeping>
jbenet: that wouldrequire changing everyones configs
<jbenet>
(2 indepndent docker procs i mean -- maybe you're onto this already, or its bad somehow)
<lgierth>
jbenet: yeah that's what i wanna do
<richardlitt>
yeah, 80 is an awesome birday. :)
<jbenet>
whyrusleeping no it would not
<whyrusleeping>
same keys?
<jbenet>
_same_ keys.
<lgierth>
jbenet: and have only one of them take the default ports
<jbenet>
lgierth: ahhh ports may screw this up
<whyrusleeping>
jbenet: yeah, it would require everyone to change their configs
<lgierth>
why change config?
<jbenet>
we can update dev04.0 configs i guess.
<jbenet>
lgierth: port is in multiaddr
<lgierth>
on half of the hosts, v03x would claim the default ports, on the other half v04x
<lgierth>
while we slowly phase out v03x, v04x claims the default ports on more hosts
<lgierth>
this will be much easier with multiaddr supporting /dns
<jbenet>
ralphtheninja yep you should say "our" :)
<ralphtheninja>
hehe
<jbenet>
lgierth: yeah indeed.
<whyrusleeping>
I think i'm up next for sprint?
<whyrusleeping>
321go:
<whyrusleeping>
- [x] libp2p example for @diasdavid
<jbenet>
lgierth: we should still have some direct ips even then, in case DNS blows up.
<whyrusleeping>
- [x] fix and test set/append data for ipfs object patch
<whyrusleeping>
- [x] cleanup mfs code a bit, refactoring how it pins
<whyrusleeping>
- [~] ipfs files flush
<whyrusleeping>
- [x] let @mafintosh convince me to go to svalbard for arcticphp
<whyrusleeping>
- [x] paris?
<whyrusleeping>
- [ ] distibutions build script nearly done, just some small cleanup to do
<lgierth>
jbenet: yep agreed
<The_8472>
ipv6!
<lgierth>
whyrusleeping: where are you now actually? :)
<The_8472>
for bootstrap too
<dignifiedquire>
whyrusleeping: distributions ships this week, otherwise I have to come to Paris enforcing it
<ralphtheninja>
whyrusleeping: arcticphp? lol
<richardlitt>
uh
<richardlitt>
I am not going to arcticphp
* richardlitt
frantically cancels tickets
<lgierth>
The_8472: yep have ipv6 on my list
<whyrusleeping>
lgierth: near the louvre
<The_8472>
\o/
<lgierth>
you mean *le* louvre?
<whyrusleeping>
THE
<whyrusleeping>
THE LOO VER UH
<dignifiedquire>
whyrusleeping: I thought it was called arcticjavascraps
<daviddias>
- [x] Copenhagen IPFS Meetup - It was great, more users of IPFS (and following close libp2p and js-ipfs) than I imagined
<daviddias>
- [x] Trip logistics
<daviddias>
- [x] Plan for ArticJS
<lgierth>
what a productivity monster
<noffle>
nice
<lgierth>
hi noffle o/
<richardlitt>
no, he's jsut good at subitizing
<noffle>
lgierth: \o
<dignifiedquire>
daviddias: is flooding ipfs activty
<jbenet>
whyrusleeping: you have to remember to post the notes from the sprint meeting in the sprint issue itself at the end, and excract todos out of that.
mec-is has joined #ipfs
<ansuz>
daviddias came and visited me and was so inspired that he got a lot done
<ansuz>
<3
<whyrusleeping>
jbenet: what?
<richardlitt>
jbenet: good catch.
<richardlitt>
daviddias: what is your star?
<daviddias>
Thank you :) However, I lost a bunch of time this week with 1) missing flight 2) losing bag 3) buying some extra cloths 4) all the flights 5) hard time getting decent Wifi (or any wifi at all) in Paris
<richardlitt>
daviddias: thanks so much for the ipfs/api help. Sorry I've been offline for the past few days. Travel has made time hard.
<mec-is>
jgraef which version of hexdump are you using? The repo needs a good requirements.txt... (:
<whyrusleeping>
ah, i see. didnt notice since richardlitt posted it
<daviddias>
Thank you :) However, I lost a bunch of time this week with 1) missing flight 2) losing bag 3) buying some extra cloths 4) all the flights 5) hard time getting decent Wifi (or any wifi at all) in Paris
<lgierth>
mec-is jgraef: hold on for a second while we're wrapping up the sprint sync
<daviddias>
richardlitt: lol..
<jbenet>
whyrusleeping: oh it is there! i was ctrl+f-ing whyrusleeping
<richardlitt>
:)
<richardlitt>
Still, should be extracted todos
<richardlitt>
I'll get better at not posting it
<whyrusleeping>
jbenet: me too
<whyrusleeping>
i've had roughly 4 hours of decent wifi this week
<richardlitt>
Oh! Btw, I have a flight later today (22:30 UTC) , so I'll miss some of the video chats this week. dignifiedquire has already agreed to set up the links for me for the ones i'll be missing. Sorry about that, thought it was later than it was.
<jbenet>
richardlitt: np thanks for the heads up
<richardlitt>
jbenet: np
<daviddias>
richardlitt: re: api. no problem, working on js-ipfs makes me go through all of these things and have a more in depth perspective of how things work, it gets easy for me to complement or review what you already had
<richardlitt>
daviddias: agreed. Thanks a ton.
<richardlitt>
You're the best.
<richardlitt>
daviddias: good work on awesome-hack-spots
<richardlitt>
there are a TON of resources for this already
<richardlitt>
but coming from an industry leader in hacking is smart.
<daviddias>
aahahah
<richardlitt>
ogd has a list of cat cafes that is also particularly good.
<daviddias>
nice
<richardlitt>
although I think ogd's list is just for Taiwan?
<lgierth>
the cat cafe in berlin is really boring
<lgierth>
who's next?
<richardlitt>
I'll make a PR with links to the other resources. This is one of my areas of expertise, I once had a startup for this. edinburghcafes.com
<jbenet>
whyrusleeping: i seem to recall important things we discussed last week for go-ipfs and 0.4.0-- i dont recall which. they dont seem to have moved forward? shipping 0.4.0 soon is important and there's still lots to get done?
boxxa has joined #ipfs
<drathir>
about 0.4 branch how much destructive it is in compare to 0.3.11-dev ?
<jbenet>
daviddias: awesome stuff with all those pieces-- sorry i didnt get to the spec on chunker stuff. do you need that this week? am still slammed with other things
<lgierth>
jbenet: yeah much of it is infrastructure :/ getting there
<whyrusleeping>
jbenet: 0.4.0 is essentially waiting on the infrastructure stuff and distributions
<brimstone>
whyrusleeping: are you going to be at hope XI?
<jbenet>
whyrusleeping daviddias NeoTeo: great stuff with the copenhagen meetup! thanks for putting that on! did people have stuff to show? (if so would be good to have writeup)
<whyrusleeping>
i havent seen any complaints about 0.3.11 breaking anything, so i can ship that tomorrow
<jbenet>
i think we should start posting on blog when meetups happen and if cool things happen feature them there
<jbenet>
whyrusleeping: i dont think so-- i seem to recall some dev stuff needing to happen still. i dont recall what exactly. maybe our notes need to be better.
<richardlitt>
jbenet: agreed about using the blog for that
<richardlitt>
jbenet: it's your turn, whenever you're ready
<daviddias>
jbenet I have more stuff on my plate to keep me busy, also I can continue working on it and make it all protobuf'i as we talked the last time, having a clear spec will essentially help me make sure what gets implemented is the right thing
<whyrusleeping>
jbenet: the only dev stuff was the deadlock stuff
<jbenet>
whyrusleeping i think it was something having to do with js maybe?
<whyrusleeping>
which has since been resolved
<jbenet>
whyrusleeping: yay! \o/
<whyrusleeping>
i think it was left off of my sprint update because githubs '1 week' excludes things done last monday
<jbenet>
whyrusleeping: maybe something having to do with ipld blocks? sigh i have a strong feeling we're forgetting something important.
<whyrusleeping>
jbenet: multicodec things?
<whyrusleeping>
i didnt really think we wanted that in dev0.4.0
<richardlitt>
to be fair: This was an exceedingly short week
<richardlitt>
Due to the Monday sprint shift
<jbenet>
whyrusleeping: btw earlier today i thought about https://github.com/ipfs/go-ipfs/issues/2185 -- your call whether it goes in-- but we have a good chance now while we break things
<jbenet>
(blake2 will improve the cpu bottlenecks)
<jbenet>
whyrusleeping: you're right-- that's not on dev0.4.0 since ipld is not ready
<jbenet>
whyrusleeping: but dont think that waas it. who knows. maybe it will come up in hangout today
<jbenet>
daviddias: ok sounds good
<jbenet>
This week I've been working through my large holiday backlog, reconnecting with various people in the IPFS periphery, and handling many start of year organizational things. Lots of email. Haven't made it to github proper yet. Sorry-- if any PRs need my pressing comments, CR, please bring them up to me via IRC or https://github.com/jbenet/todo
<jbenet>
- [x] whittled backlog
<jbenet>
- [x] lots of IPFS related email
<jbenet>
- [x] worked on PRs with @richardlitt, discussed various community things
<jbenet>
- [~] IPLD spec work with @mildred -- i still need to do more here this.
<jbenet>
- [ ] finalize IPLD
<jbenet>
- [ ] latest vs stable webui comments (@jbenet) in ipfs/webui#167 -- didnt get to this
<jbenet>
- [x] lots of protocol labs org things
<jbenet>
EOF
<richardlitt>
L)
<richardlitt>
*:)
<whyrusleeping>
jbenet: hrm... blake2 might be nice. but we should probably do it at the same time that we break everything else for ipld
<jbenet>
yeah i guess ipld will also be backwards "problematic"
<whyrusleeping>
yeah... thats one word for it
<whyrusleeping>
lol
<jgraef>
mec-is. I'm back. I didn't want to publish it yet ;)
<jgraef>
One sec, I'll include the hexdump stuff
grahamperrin has joined #ipfs
<The_8472>
<jbenet> (blake2 will improve the cpu bottlenecks) <- shouldn't bulk hashing be trivially parallelizable?
<richardlitt>
Thanks for all of the help this week, @jbenet. Glad to have those PRs merged
<The_8472>
do you run NVMe raids on atom CPUs or something that hashing is the bottleneck? ^^
<jbenet>
The_8472 in some cases CPU hashing is the bottle neck (low power devices) -- for them, blake2 will make a big difference
<jbenet>
yeah some atoms :)
<jbenet>
richardlitt: happy to
<The_8472>
yeah, but isn't their IO slower too?
<jbenet>
whyrusleeping: it wont be a hard break, only a fwd break (like new content wont be readable by old nodes, but they can still speak
<whyrusleeping>
hashing is the bottleneck on my laptop. sha256 ~ 200MB/s, my hard drive is around 800MB/s
<richardlitt>
noffle: Your turn?
<jbenet>
we will want the damn -Wall
<The_8472>
whyrusleeping, per core?
<whyrusleeping>
The_8472: its parallelizable, but not trivially
<lgierth>
jbenet: -Wall?
<whyrusleeping>
note that we're building a merkletree
<jbenet>
and we will want autoupdate :]
<The_8472>
sure, but leafs take up a lot of time
<jgraef>
mec-is, pushed.
<jbenet>
yeah in large trees leaves will dominate
<ralphtheninja>
richardlitt: I'm guessing you're the sprint master?
<The_8472>
but you're right, power efficiency is always nice
<jbenet>
ralphtheninja: he is
<noffle>
not much for me yet. "I've just been trying to ramp up on everything (reading lots of specs over the holidays!), and figure out what I'm going to be focusing my efforts on for Q1. I expect to be bugging you all a lot over the coming weeks!"
<richardlitt>
ralphtheninja: yep.
<richardlitt>
ralphtheninja: I think it says that in the long Readme, too.
<ralphtheninja>
richardlitt: I should have read that before, reading now though :)
<richardlitt>
noffle: that's still awesome
<richardlitt>
super excited
<richardlitt>
thanks.
<richardlitt>
Does anyone else want to go?
s_kunk has quit [Ping timeout: 240 seconds]
<kyledrake>
I can give a quick update
Matoro has quit [Ping timeout: 255 seconds]
jhulten has joined #ipfs
<jbenet>
kyledrake: post it to the pm sprint first pls?
<jbenet>
richardlitt: did you see what i meant about the number?
<jbenet>
(on blog)
<richardlitt>
jbenet: no, I didn't
<richardlitt>
I'm sorry.
<richardlitt>
jbenet: can you explain it further?
cryptix has joined #ipfs
<whyrusleeping>
gonna try and run to a bigger coffee shop with better internet
<dignifiedquire>
jbenet: need to review & fix up the copy for the distributions page before we ship it, do you think you have some time later for that?
<dignifiedquire>
or should I just do a first pass and then ping you to review?
<jbenet>
whyrusleeping: ok sgtm. you're in paris or portugal?
<richardlitt>
Hmm. `#` must be the comment
<richardlitt>
I'll change it, pushing a PR in two seconds
<dignifiedquire>
oO this is not on ipfs.io/lbog
<jbenet>
dignifiedquire: the copy (i.e. the text, right?) -- maybe richardlitt can help too?
<jbenet>
dignifiedquire: not yet, will be soon
<jbenet>
and dignifiedquire: just poke me here with things to CR this week
<dignifiedquire>
jbenet: yes the text, and details like the copyright in the footer
<dignifiedquire>
jbenet: okay will do
jhulten has quit [Ping timeout: 240 seconds]
<The_8472>
<jbenet> whyrusleeping: it wont be a hard break, only a fwd break (like new content wont be readable by old nodes, but they can still speak <- you will have to think about that eventually
<The_8472>
hashes will have to be deprecated when they become vulnerable to new attacks
wiedi has joined #ipfs
<jbenet>
i think the copyright should say "Protocol Labs" for now -- im looking into making a formal org we can assign copyright to (like "IPFS Team" or "IPFS Foundation" -- but foundations are trickier)
<richardlitt>
dignifiedquire: happy to CR the copy - can you send me the link?
<Kubuxu>
jbenet: have you thought about LLC?
<dignifiedquire>
richardlitt: not written it yet
<Kubuxu>
I suppose you are US based.
<jbenet>
The_8472 indeed- multihash helps but not a full solution. if you have good ideas to make it upgradeable im all ears. btw this may be a lot easier with the index btrask and i came up with -- https://github.com/ipfs/notes/issues/89
<richardlitt>
dignifiedquire: cool. When you do, ping me
<lgierth>
ok gonna hunt down something to eat, see you in 0h50m for infrastructure hangout
<jbenet>
there wont be sub versions, we'll get to 42.0 43.0 44.0 .... 111.0
<jbenet>
i'll merge this as is, we can change it later
<jbenet>
as long as the links dont change--
<M-mubot>
as long as the links dont change has -1 points
<kyledrake>
I was on holiday break the last few weeks. I'm in San Francisco this week meeting with people. I'm starting a new sprint. First agenda is to dust off the 0.4.0 blog post, update things and send it upstream.
<Ape>
What is that M-mubot point system?
zorglub27 has quit [Read error: Connection reset by peer]
<kyledrake>
I'm done
<richardlitt>
kyledrake: cool
<richardlitt>
thanks. :)
<richardlitt>
looking forward to blog movement
Matoro has joined #ipfs
<jbenet>
kyledrake are you there through next week?
<kyledrake>
jbenet until Tues very late.
<kyledrake>
I brought the rain with me. Sorry mappum
<richardlitt>
ok! The sprint sync is officially closed.
<richardlitt>
See some of you in 40 mins for the first talk
<dignifiedquire>
40?
<richardlitt>
right?
* richardlitt
checks sched
<dignifiedquire>
10min
<richardlitt>
hmm. 10
<richardlitt>
:D
<richardlitt>
apps on IPFS in ten
<jbenet>
ooof 10
<jbenet>
we took a long time this time
<richardlitt>
jbenet and dignifiedquire: please prepare appropriately
<The_8472>
jbenet, hrrm, alternative hash lookups can help with rewritten data, but it won't fix references using broken hashes. they're inherently untrustworthy.
<jbenet>
hey noffle able to attent the apps hangout in 10? want to talk about ipget a bit
<The_8472>
so maybe bake an optional "lookup with legacy cruft" support into apis?
<noffle>
jbenet: can do!
<mec-is>
jgraef pull request sent, see what you can merge
ygrek has quit [Ping timeout: 272 seconds]
<dignifiedquire>
jbenet: should I keep distributions on apps or should we move it maybe to infrastructure?
akkad has quit [Excess Flood]
<jbenet>
The_8472 hmm something like that. i suspect there is an elegant solution to this nasty problem.
akkad has joined #ipfs
<jbenet>
dignifiedquire as you wish but i cant make it to infra
<dignifiedquire>
then I'll keep it in apps for now
<dignifiedquire>
whyrusleeping: you got wifi again?
<jbenet>
sgtm
akkad has quit [Excess Flood]
akkad has joined #ipfs
* whyrusleeping
back
<jbenet>
richardlitt can we move the "## participants" line to above "### lead" in the template?
<jbenet>
it always confuses me
<ralphtheninja>
afk walking dog .. I'll jump on some of the streams later when I'm back
akkad has quit [Excess Flood]
<noffle>
jbenet: still figuring out how to get into this hangout (oh the irony)
<dignifiedquire>
welcome back whyrusleeping you wanna join us in a couple of minutes for apps hangout, will be talking about distributions
<ralphtheninja>
are the hangouts recorded btw?
<The_8472>
jbenet, hrm, even with certificates their best solution is to add time and verify that the time has sane values at time of issuing. and verifying time in distributed environments is tricky too
<richardlitt>
jbenet: no problem
joshbuddy has joined #ipfs
* drathir
also also wonder if is considered adaptate or maybe even it presentt some kind of usage ipfs as private secure cloud storage system between multiple devices?
<The_8472>
and checking "is it really old" only helps against collision attacks
<jbenet>
noffle: hangout hasnt been posted i dont think? (oh are there old links?)
<jgraef>
mec-is, I'm on it.
<dignifiedquire>
ralphtheninja: yes, but not publically available
<jbenet>
richardlitt: might be good to post links 2-3 mins before so that we can try to start on time
<Ape>
Can I watch the hangout somewhere?
<dignifiedquire>
jbenet: richardlitt is just setting them up
<ralphtheninja>
kk maybe I can ask for a link later
<richardlitt>
setting it up right now
<dignifiedquire>
Ape: yes youtube link incoming soon
jaboja has quit [Ping timeout: 276 seconds]
<The_8472>
drathir, you can obviously encrypt your payload
<jbenet>
The_8472 yeah i was thinking of some attestation based thing-- there may be a protocol to undertake before (ike the root node points to an object that links to public keys to be trusted through cryptographic breaks
<jbenet>
but in that case might as well use ipns
<jgraef>
mec-is, Those unit tests don't assert anything.
<The_8472>
a solution for hashed static content would be nice. but it kinda seems intractable. if your hash is broken (preimage) you're hosed. because you can never be sure what it really should have been. any input that validates is equal
<The_8472>
and with collisions you still get forgeries
<drathir>
The_8472: yes but somethin like eg. auto encrypted pool by gpg keys adding to it or somethin like that? somethin at protocol lvl which be good alternative to owncloud seafile or even dropbox like services?
<The_8472>
drathir... idk... could enc2fs be layered ontop of fuse?
<The_8472>
enc2fs being a fuse thing
<The_8472>
inception.jpg
<jgraef>
mec-is, And I guess tearDown should not be marked with @unittest.skip("debug"), right?
captain_morgan has joined #ipfs
<richardlitt>
the chat is being delayed a few seconds
<richardlitt>
Will be starting soon
<drathir>
or something like auto clone chosen user added resources to own database... btw thats only my probably crazy one loud thinkin ideas of usage...
akkad has joined #ipfs
<Kubuxu>
Trolls gonna troll
<Kubuxu>
hahha
<Kubuxu>
You can use time slider
<whyrusleeping>
ansuz: hows that private etherpad thing going?
<Akaibu>
Yea, this asshole needs to get banned or something if that's possible
grahamperrin has left #ipfs ["Leaving"]
ygrek has joined #ipfs
Matoro has quit [Ping timeout: 256 seconds]
<jgraef>
mec-is, merged. Thanks for writing the unittest code :)
<jgraef>
I'll write tests when I'm done implementing the API.
warner` is now known as warner
cryptix has joined #ipfs
<ansuz>
who's trolling?
<ansuz>
whyrusleeping: it's going well, I think I'm about ready to close all the open issues and release
<ansuz>
mind you, that's for the xwiki integration, not the standalone version
<jgraef>
mec-is, Isn't the requirement like all there is for the tests? :D For the API, take a look in ipfs.api.* where * is a ipfs command (everything except proxy)
moreati has quit [Ping timeout: 272 seconds]
<mec-is>
jgraef once you have enough tests you can ninja coding whatever I think
<mec-is>
jgraef yep but if you write it in natural language into the docstring is easier to work wiyh the codebase
<ipfsbot>
[js-ipfs] diasdavid pushed 1 new commit to master: http://git.io/vzJxu
<ipfsbot>
js-ipfs/master dbe8d7c David Dias: Release v0.0.2.
elima has joined #ipfs
<mec-is>
jgraef for tests are called 'requirements' the things that the test has to perform, so if you give me some reqs i can write the test cases. This can be a good way to work on the repo
<Kubuxu>
OVH has hosting in US and good backbone back to Europe (1Gi) for cheaper dedis (10Gi) for better ones. Also there is cheaper branch SoYouStart (250Mi to the Internet, 1Gi inside). JFYI
<jgraef>
mec-is, I mean you can check for all functions, if they return correct values.
joshbuddy has quit [Quit: joshbuddy]
<mec-is>
jgraef ... like you write reqs along with the implementation, and then you pass them to me and i add to the unit. Sounds good?
<mec-is>
jgraef yep, but you have to write me what they are supposed to return... edge cases etc...
<jgraef>
mec-is, well. I can do that.
<mec-is>
jgraef i will check also by myself, but if you write them down it's far more smooth
<mec-is>
Cooool man
<jgraef>
mec-is, btw. pull. I just implemented pin and name a few minutes ago.
cryptix has joined #ipfs
<mec-is>
Ok, i go for dinner now and come back later
jaboja64 has joined #ipfs
<Kubuxu>
lgierth: I have (not so)funny sorry. I finally fixed my dedicated, what I learned from that: If you have Linux with random hags/crashes, check if intel-ucode is installed. Just sharing as it took month to track it down (I didn't really tracked it down, just run out of ideas long time ago and I was reinstalling Arch on my notebook).
<Kubuxu>
During which I was reminded by Arch wiki to install intel-ucode :)
<lgierth>
hehe thanks
<lgierth>
my experience
<lgierth>
with ovh hasn't been too good
<lgierth>
the ui is disturbingly annoying and the payment process is just ridiculous
<lgierth>
having to manually renew every month and reenter the credit card doesn't work for us
ashark has joined #ipfs
jaboja has quit [Ping timeout: 256 seconds]
<dignifiedquire>
whyrusleeping: also the "id" for go-ipfs is just "ipfs" which is wrong, and makes problems for me, is that easy to fix?
<Kubuxu>
You can renew it for longer time, but I can see that card issue might be problematic.
<drathir>
lgierth: their not storage that data from security reasons?
<drathir>
lgierth: how aboyt bank sheulded pay for ovh assigned bank account also not supported?
<lgierth>
drathir: no they're just sketchy imo :)
cryptix_ has joined #ipfs
<Kubuxu>
Were you using main branch or the lower ones?
<lgierth>
i like hetzner -- they've been weird with their network scan policy, but reliably so, and the process around it wasn't too bad
<lgierth>
just plain ovh
<lgierth>
the smallest storage host they offer
<drathir>
lgierth: bc really month by month doin that could be annoying a little ^^ but from second side harder to forgot about that... ;p
dysbulic has quit [Quit: ChatZilla 0.9.92 [Firefox 43.0.4/20160106234723]]
<lgierth>
no, eventually you *do* forget and then your server goes down
<Kubuxu>
I had problem with Hetzner (hardware failure they told us 3 times they fixed) yet memtest was still crashing.
<drathir>
lgierth: hetzner is fine excluding hw fails and blocked/banned hetzner ip range...
cryptix has quit [Ping timeout: 250 seconds]
<Kubuxu>
OVH sends you at least 3 emails before they cut you off.
<drathir>
Kubuxu: at hetzner mostly hw fails by death of hw ^^
<drathir>
no possible to warn, but they smothly fixing that and ofc not their fault...
<drathir>
hetzner also have if good remember really nice colo offers too...
jhulten has joined #ipfs
<lgierth>
whyrusleeping: about raid 5 and parity 1 -- my intention with that was to rather to proper failure tolerance on the ipfs layer
kahiru has joined #ipfs
<lgierth>
whyrusleeping: i.e. maxing out individual hosts, and adding more to increase parity
<whyrusleeping>
right
<whyrusleeping>
thats fine
<lgierth>
that's also why i don't by usb3 raid cases anymore :)
<Kubuxu>
drathir: It was so annoying, we were even thinking it was conspiracy so we drop the lease as we had great offer at the time and they changed prices multiple times since we bought it.
<Kubuxu>
But those were just jokes.
elima has quit [Ping timeout: 240 seconds]
jhulten has quit [Ping timeout: 240 seconds]
<Kubuxu>
Wooo, gateway just saved file I had no other option to get.
<lgierth>
Kubuxu: were you vyl in the hangout?
<Kubuxu>
lgierth: no I was only on the pad. I don't know how I ended up being signed on it.
<Kubuxu>
I was mostly eating and fighting with troll.
joshbuddy has joined #ipfs
<lgierth>
which troll?
<Kubuxu>
whyrusleeping: Where should I write issue about mounting only /ipfs/ without /ipns/ (or /ipns/ in read only mode)? ipfs/ipfs or ipfs/go-ipfs
<lgierth>
nevermind
<Kubuxu>
There was troll on the Apps' pad.
boxxa has quit [Quit: Connection closed for inactivity]
<whyrusleeping>
Kubuxu: put it in go-ipfs
<Kubuxu>
kk
<noffle>
daviddias: just tried jsipfs -- this is so cool :D nice!
<daviddias>
:D :D thank you for checking it out!
<Ape>
lgierth: How are ipv6 fc00::/8 addresses useful for ipfs bootstrapping? Aren't they private local addresses?
computerfreak has quit [Quit: Leaving.]
<jgraef>
mec-is, I sent you a pastebin with some requirements. Didn't write mich text though. I hope it is self-explanatory.
<Kubuxu>
Ape: no they are Hyperboria addresses.
<lgierth>
Ape: cjdns
<Kubuxu>
local addresses are fd00::/8
<Ape>
Ah, thanks
<lgierth>
strictly speaking fc00::/8 is also private local
<lgierth>
it's just that it isn't used anywhere except for cjdns
<daviddias>
noffle: let's make sure we talk about all things js-ipfs
<lgierth>
fc00::/7 is unique local unicast, which is fc00::/8 + fd00::/8
<noffle>
daviddias: I'll be at sync :)
<Kubuxu>
AFAIK fc00::/8 is local but it usage is not defined. (it was left out just in case we screw something up in fd00::/8).
<drathir>
ralphtheninja: bc workin fine in mplayer for me...
<noffle>
ralphtheninja: ditto
<noffle>
ralphtheninja: what mplayer version?
<ralphtheninja>
2.0-728-g2c378c7-4+b1
* drathir
MPlayer SVN-r37379 (C) 2000-2015 MPlayer Team
<whyrusleeping>
kahiru: the issue is that ipfs chunks the files it adds, so the data isnt the same
<whyrusleeping>
because of that, a reflink won't work as expected
<noffle>
no errors? it has mpg123 decoder?
<kahiru>
that's a bummer
<noffle>
does it play if you DL locally first?
<Ape>
And the file would have to be on the same filesystem
<ralphtheninja>
noffle: let me check
<kahiru>
still it would be great if ipfs could be used without duplicating the data
<Ape>
Instead you could add the file to IPFS and replace the original with a symlink
<The_8472>
<whyrusleeping> kahiru: the issue is that ipfs chunks the files it adds, so the data isnt the same <- that could be stored as is on disk and chunked while reading
<whyrusleeping>
The_8472: mmm, yeah. we could eflink the file into somewhere inside .ipfs, then have some special datastore format that knows what offsets to read
<The_8472>
yup
<whyrusleeping>
we then have to code in logic for detecting COW filesystems
<drathir>
kahiru: if data is exacly the same it isnt linking to present version?
<The_8472>
avoids duplicate storage but also avoids accidental modification
<whyrusleeping>
support both that and non-cow
<whyrusleeping>
and windows
<Ape>
whyrusleeping: Just see if reflink is supported by the fs
<The_8472>
windows does reflink too i think. whatever shadow copies do
<The_8472>
not sure if they expose an API for that
<Kubuxu>
I hope all btrfs users disabled COW on .ipfs dir.
<drathir>
somethin like hash calculate before save...
<ralphtheninja>
mplayer can play it when cached locally
<Ape>
Kubuxu: why?
<kahiru>
drathir, I'm not sure I understand what you meant by that
<dignifiedquire>
jbenet: also do you have specific priorites for me to work on besides what I'm doing currently?
<Kubuxu>
Ape: because btrfs will copy the DB all the time and block are immutable either way so you don't need btrfs.
<drathir>
kahiru: some kind if even first save at disk and after checkin a hash of file if find duplicate entry deleting old one and linkin to new one but the most optimal would be checkin hash before sedownload and if present in database only linkin not downloading...
<jbenet>
dignifiedquire: shipping distributions is big for 0.4.0. then, API review. maybe improving the webui? (bugfixes / files api?) or maybe shipping station? -- not sure, other things you're interested in?
<Kubuxu>
also disabling COW disables hashing iirc
disgusting_wall has joined #ipfs
<kahiru>
drathir, oh, I see
<daviddias>
is anyone else looking into joining the libp2p sync today?
<lgierth>
yep
<Ape>
Kubuxu: But COW isn't really a problem is it? It just does a bit wasted hashing. If the data is immutable there wont be much fragmentation
<dignifiedquire>
jbenet: looking into wikipedia mirroring but we think that for that we first should have a better story for clustering, as the data is just massive
Senji has quit [Ping timeout: 264 seconds]
<dignifiedquire>
so that's pretty much what I'm interested in at the moment (what you listed)
<dignifiedquire>
at some point I want to help daviddias a JS ipfs
<dignifiedquire>
*in
<Kubuxu>
Ape: there is leveldb which might get much bigger than my 2M (only about 200Mi of pinned files and fresh after GC).
<dignifiedquire>
but for that I need more time in a block
* The_8472
doesn't understand a word
<Ape>
Kubuxu: I don't think it would be any problem in any case
<Kubuxu>
In case of much bigger DB it might be a problem.
<Kubuxu>
And as disabling COW in case like that looses you nothing...
<achin>
richardlitt: my proposal: each monday, whoever is able to start working on the weekly round up first should open an issue on the weekly repo indicating that. this will serve as a signal to others to not duplicate effort
<whyrusleeping>
ping
<drathir>
hash search in db shouldnt be so hard/long to verify, longer is generate the hash and waitin for that before download the file...
<drathir>
Kubuxu: but anyway save a lot of bandwidth on big files...
<dignifiedquire>
whyrusleeping: who are you pinging?
<drathir>
in theory ofc...
wiedi has quit [Quit: ^C]
<Kubuxu>
drathir: it isn't about searching hashes, it is about adding data to the DB, it will reallocate multiple HDD blocks in case of COW.
<Kubuxu>
and snapshoting sill works as it just tags blocks as CoW and after they are once CoWed then there is no CoW.
<whyrusleeping>
dignifiedquire: the internet
<whyrusleeping>
its being dumb for me right now
<The_8472>
CoW shouldn't be an issue though. if it's async it goes through the page cache. if it's sync it can go in the bcache if you have it in writeback mode
<NeoTeo>
jbenet: The meetup was my pleasure :) We had a very informal but thorough explanation of the IPFS concept by whyrusleeping and daviddias which quickly turned into an hours long Q&A. Everyone there was enthused and keen on trying it out. I'm thinking next meetup will be an 0.4.0 install party ;)
<Kubuxu>
While adding big files you already have a lot of HDD i/o. If you can remove some, without loosing almost anything, why not?
<The_8472>
websockets might be simpler than webrtc?
<NeoTeo>
jbenet: also, I want to make sure you didn't miss my msg of a week (or two) ago about arranging to pay you back for your 32C3 ticket :)
<The_8472>
yeah, addon is probably more efficient. ipfs instance per page would be insane
<The_8472>
or even per domain (service workers)
<Kubuxu>
I proposed that on FF's add-on GH.
<Kubuxu>
Coffee shop paid by an hour... interesting.
hellertime has quit [Quit: Leaving.]
jgraef has joined #ipfs
<noffle>
daviddias: once we have js-ipfs nodes working in the browser, I wonder if we could use something like https://github.com/substack/page-bus to let users run an ipfs node (heavy!) in one tab and reuse it across any other tabs that want an ipfs node quick
<noffle>
same domain only though, so :sadface:
<The_8472>
there already are broadcastchannels and service worker for that purpose, no need for an external lib
<daviddias>
There is an mdns API for the browser now
<daviddias>
We could find the other node in the other tab and WebRTC into it
<noffle>
oh interesting
<The_8472>
but then you'd still be tabcentric and one domain would be at the mercy of another one. seems unclean
<The_8472>
addon injecting a shared instance seems safer
<noffle>
The_8472: yes, that'd be ideal
<drathir>
noffle: question if that keepp correct data separations, but interesting idea...
* lgierth
very much likes the plugin thing
<lgierth>
eeh addon
<lgierth>
might make it easier for browser or distributions to ship ipfs by default
<lgierth>
also the firefox addon just hit 1.4.0 yesterday and supports fs: now
rami_ has quit [Quit: Page closed]
jgraef has quit [Quit: Bye]
<richardlitt>
achin: Sounds good to me.
libman has joined #ipfs
m0ns00n has quit [Quit: undefined]
wiedi has joined #ipfs
<richardlitt>
jbenet, achin: woo! That looks great.
Encrypt has joined #ipfs
<drathir>
lgierth: 1.3.1 still in addon database...
Shibe has joined #ipfs
<achin>
that's a good lookin' picture of whyrusleeping at ccc!
<whyrusleeping>
Huh
<whyrusleeping>
?
<lgierth>
drathir: yeah it got tagged yesterday, reviewed today, and probably will hit the database tonight or tomorrow i guess
<Kubuxu>
lgierth: also note, I've learned today that utp multiaddr is: /ip6/[ip]/udp/4001/utp/
<lgierth>
yeah
<The_8472>
isn't that a bit of a misnomer?
<drathir>
thats sounds little strange ^^ 22:41 < achin> that's a good lookin' picture of whyrusleeping at ccc!
mildred has joined #ipfs
<Kubuxu>
The_8472: no as utp is build on top of upd
<Kubuxu>
udp
<The_8472>
i know that
<The_8472>
but there are other udp based transport
<The_8472>
*transports
<whyrusleeping>
Yes
<whyrusleeping>
Udt, quic
<whyrusleeping>
Sctp
<The_8472>
yeah, so the multiaddr needs fixing
<whyrusleeping>
Why
<The_8472>
wrong layer
<lgierth>
why?
<whyrusleeping>
How so?
<lgierth>
it's utp-over-udp-over-ip
<lgierth>
can't talk utp straight over ip
<The_8472>
tcp denotes tcp-over-ip, udp doesn't denote udp-over-ip but utp-over-idp-over-ip