alexandria-devon has quit [Read error: Connection reset by peer]
<krl>
jbenet: ping
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
mildred has joined #ipfs
alexandria-devon has joined #ipfs
<krl>
jbenet: when you're back, you could try out the rebrowserify branch, and see if make publish gets you QmS2HL9v5YeKgQkkWMvs1EMnFtUowTEdFfSSeMT4pos1e6
<krl>
webpack is broken, wip
<krl>
i would personally prefer to use some devtool that actually does browserify, and is less framework-ey
<krl>
should i be using io.js? i have not kept up with the schism
<jbenet>
I think it's merging back soon
<krl>
substack: browserify should be deterministic right?
<tperson>
I haven't really kept up with everything either, but, ya that ^^
<jbenet>
ffs untaring and compiling node takes longer than downloading a build binary and installing it
<substack>
krl: unless transforms or plugins do something non-deterministic
<substack>
there's a whole phase of the browserify pipeline that topologically sorts the packages to preserve determinism
<krl>
awesome
<substack>
and a number of unit tests too
<krl>
that really fits our usecase
<substack>
really hard to have verified and audited builds without that
<krl>
i hope less also gets that right
<jbenet>
krl: i get QmR1LubbsYWVBe83x9TRdSozSZ2x3ArP1cnmhvWtnH1BA2
<jbenet>
(deterministically)
<krl>
jbenet: could you provide?
<jbenet>
krl: let's avoid the `npm install` step if it;s been run before (any way to check this? maybe depend on the package.json file (i.e. if its updated it runs npm install)
<jbenet>
krl: i think im behind a pretty evil nat atm
<jbenet>
sec
<krl>
jbenet: i'll have to clean up the Makefile anyway, sounds like a good idea
<tperson>
This is the rebrowserify branch correct?
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<krl>
tperson: yes, but rebrowesrify
<krl>
-_-
<tperson>
What hash do you get on that branch krl?
<krl>
is that if we're planning on splitting up unixfs dirs to support thousands of links
<krl>
this will break the path scheme mapping between mdag - unixfs
<kyledrake>
jbenet whyrusleeping no useful stack dumps yet, but here's the general pattern. I turn on ipfs and start the archive workers, and it speeds up again. After about an hour it starts to get slower and starts to gums up, with no obvious bottlenecks. I have a script restart both of them every hour, and things go back to normal. You mentioned that you need to
<kyledrake>
restart the gateway.ipfs.io servers sometimes, so I'm thinking it's maybe the same state issue
<jbenet>
krl: kk sec
<jbenet>
kyledrake: yeah i think so too
<kyledrake>
jbenet whyrusleeping so ipfs add -r everything is actually worse than my process to add one site at a time, because it will basically get to some point and then stop. I think you could reproduce this pretty easily by just trying to drop half a million random small files into it with an ipfs add -r.
<jbenet>
krl: what's your addr? im having a hard time connecting to one of my ndoes outside--
<krl>
or try: /ip4/131.72.139.47/tcp/4001/ipfs/QmU7pJf8BpY24gAKB92Xm1whcmV5FNKzPCbExKwYG6HYbi
<kyledrake>
like, even if you just wrote a script to create 500k random uuid files filled with /dev/urandom with 4k-200k of data per file. You probably don't even need that many, the ipfs add -r everything stopped at around 25k last time I ran it.
<kyledrake>
If that doesn't work let me know and I'll figure out how to send a tarball over to you guys
<jbenet>
krl: something weird's going on-- may be the network im in
<krl>
jbenet: tar it up and post it somewere? :/
<jbenet>
krl: provided
<krl>
ok got it
<kyledrake>
If you'd like me to I can probably take care of that script and try to repro.
<jbenet>
kyledrake: that'd be great actually
<kyledrake>
cool I'll do that tomorrow
<jbenet>
kyledrake: if we make a test case we can add it to the test suite and make every commit pass it.
<krl>
jbenet: huh
<krl>
diff says: Only in jbenet2/static: bootstrap
notduncansmith has joined #ipfs
<jbenet>
hmm
notduncansmith has quit [Read error: Connection reset by peer]
<krl>
you have an extra directory there for some reason
<krl>
well, a tree of empty dirs
<jbenet>
krl: removed it-- not sure why git status wasnt picking it up.
<jbenet>
(not ignored)
<jbenet>
krl: yay got QmS2HL9v5YeKgQkkWMvs1EMnFtUowTEdFfSSeMT4pos1e6 \o/
<krl>
woho!
<krl>
ok, next i'll clean up the Makefile and see if i can get a hot-reload dev env going
<krl>
but that's another day
neoteo has joined #ipfs
<jbenet>
krl: sounds good-- shall we publish this?
<jbenet>
want to push this commit to 0.2.0 ?
Wallacoloo has quit [Quit: Leaving.]
<krl>
jbenet: there's no working dev-env in this commit
<krl>
but it would be nice to have the webui upgrade out finally
<jbenet>
krl: ah
<jbenet>
probably ok i think?
<jbenet>
yeah either way.
<krl>
yeah, users won't notice :)
<jbenet>
im fine waiting too-- could cut releases for both tomorrow.
<jbenet>
whatever you prefer :)
<krl>
i might not have much time tomorrow before the sprint
<cryptix>
np :) can you roughly summarize what it is about?
<cryptix>
i guess you can - just if you want to..
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
Evermore_ has joined #ipfs
Evermore has quit [Ping timeout: 250 seconds]
<daviddias>
for sure :) roughly roughly is to create a way to access the untapped amount of resources available on end user machines for resource sharing/volunteered computing
compleatang has quit [Quit: Leaving.]
<cryptix>
ah like seti or fold@home?
<daviddias>
exactly, but fully decentralized (using a DHT) and using the browser (most ubiquitous platform)
<cryptix>
awesome stuff
<cryptix>
thats really cool
<daviddias>
thank you :)
<daviddias>
I did managed to get speed ups (using RayTracing jobs to do performance tests), however I do know that I only reached the tip of the iceberg, until we have things such as WebRTC inside WebWorkers (to enable parallelisation of routing)
<cryptix>
my mind stumbled over webgl to accelarate specific workloads - but i'm not sure the webgl apis are sufficient for cuda-like stuff
compleatang has joined #ipfs
<daviddias>
I haven't used WebGL yet, for the type of jobs I was running, when partitioned enough, each task would take max 50ms, it was the transport+routing that took the majority of the time
chrisr_ has left #ipfs [#ipfs]
<cryptix>
oh yea. i see, of course. without central instancesm, the coordination gets non trivial really fast I assume
ashleyis has joined #ipfs
* cryptix
has to go (traiinnzzz)
<cryptix>
l8ters
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
warner` has joined #ipfs
hellertime has joined #ipfs
<daviddias>
yes, exactly :) . ok later cryptix
headbite1 has joined #ipfs
_fil__ has joined #ipfs
wking_ has joined #ipfs
headbite has quit [Ping timeout: 258 seconds]
_fil_ has quit [Ping timeout: 258 seconds]
wking has quit [Ping timeout: 258 seconds]
sprintbot has quit [Ping timeout: 258 seconds]
warner has quit [Ping timeout: 258 seconds]
lachenmayer has quit [Excess Flood]
HastaJun has quit [Ping timeout: 258 seconds]
brab has joined #ipfs
lachenmayer has joined #ipfs
inconshr_ has quit [Remote host closed the connection]
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<ralphtheninja>
so I'm playing around with 'ipfs dht put/get'
<ralphtheninja>
it works locally but if I try to get a key from another peer I get 'error: dial attempt failed: peer has no addresses'
JasonWoof has joined #ipfs
<JasonWoof>
is there a copy of your paper ("ipfs-p2p-file-system.pdf") where the paragraphs don't get scrambled when I select/copy them?
<JasonWoof>
I'd like to send the text through espeak, but the PDF doesn't have the text in the correct order on lots of pages (try selecting the left column on page two for example)
<lgierth>
krl: i'll join you for the p2p meetup -- can we take the sprint meeting from there again?
<cryptix>
JasonWoof: i'm sure jbenet can give you the raw text - just stick around until he wakes up :)
<ralphtheninja>
is christofer (the swede living in berlin) in here? :)
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<cryptix>
ralphtheninja: krl <- :)
Leer10 has quit [Ping timeout: 256 seconds]
<ralphtheninja>
cryptix: thanks :)
nessence has quit [Remote host closed the connection]
headbite1 is now known as headbite
brab` has quit [Ping timeout: 246 seconds]
brab` has joined #ipfs
<JasonWoof>
thanks cryptix lgierth :)
<JasonWoof>
a quick regex to delete \something and it reads well :)
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
brab` has left #ipfs ["ERC (IRC client for Emacs 24.5.1)"]
<cryptix>
JasonWoof: awesome :) quite sad to see how much of a wasteland latex typesetting still is
<cryptix>
espc wrt its age
<JasonWoof>
I just watched the demo again, mostly seems awesome, except for the one horrifying thing: the webui seems to be a non-privilaged app (served via ipfs) and yet it displays the private key of the demo dude
<cryptix>
JasonWoof: that goes over the :5001 api, its not unrestricted
<JasonWoof>
cryptix: so, js apps loaded over that port have a different kind of access to local stuff?
<cryptix>
we currently whitelist sepcific hashes
<cryptix>
(and krl added an --unrestricted option to allow for easier local dev)
<JasonWoof>
ok
<JasonWoof>
I didn't think I would be the first to notice that... but it is a horrifying aspect of the demo, since he doesn't mention why it's ok
pfraze has joined #ipfs
<cryptix>
thats a good point - jbenet o/
<cryptix>
jbenet whyrusleeping: dont know how you guys feel about this but i'd like to try a different solution for sprint meetings (maybe mumble) i had strange audio artifacts the last times which made it really hard to understand everything - i dont need hd video streams if i dont understand you guys :-/
<cryptix>
but it can wait for input from others cc wking_ dPow krl
therealplato has joined #ipfs
<ipfsbot>
[go-ipfs] cryptix force-pushed feat/httpApiTcpPort0 from 563760b to cea5aa2: http://git.io/vkWm3
<ipfsbot>
go-ipfs/feat/httpApiTcpPort0 cea5aa2 Henry: http endpoints: dont print before listen...
<cryptix>
pew pew everything done for this sprint :]
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
wking_ is now known as wking
HastaJun has joined #ipfs
<daviddias>
^5 cryptix :)
<whyrusleeping>
gooood morning everyone
<whyrusleeping>
cryptix: we can try talky
<whyrusleeping>
its supposedly p2p
<cryptix>
daviddias: :) thx
<cryptix>
whyrusleeping: hi :)
<cryptix>
the thing i fear about truly p2p solutions that slow uplinks will just burn their upstream because of no real multicast
<cryptix>
asymetric links are still a major pain in .de and i belive thats also true for most countries
vijayee__ has joined #ipfs
<cryptix>
oh - i just noticed we split the travis tests into go and sharness runners. <3 to whomever did that
<daviddias>
talky might work for our scenario, since it fires a bridge for more than 5 peers are connected
<ralphtheninja>
I created an ipfs node at ipfs-1.ralphtheninja.info
<daviddias>
however, yes, no MultiCast support is still a PITA and it help improve things a lot
<cryptix>
iirc SCTP does some kind of multipath - could that be why they chose it for webrtc? sctp has been around for ages mostly unused, i think
<JasonWoof>
in some use cases, you can work around slow uplinks by having slow+steady uploading for long enough that it catches up to the (usually bursty) downloading
<whyrusleeping>
ralphtheninja: cool, welcome to the network :)
<ralphtheninja>
whyrusleeping: thanks :)
<JasonWoof>
maybe this doesn't work for video conferencing though
<JasonWoof>
seems to work well for sharing of popular files though, eg bittorrent
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<cryptix>
whyrusleeping: dude - that findProv rework really made things snappy
<cryptix>
haha oh ralphtheninja's node is also on digitalocean amsterdam.. that might explain how insanly fast that was
<cryptix>
still lookups are intense now
<whyrusleeping>
cryptix: did it finally merge?
pfraze has quit [Remote host closed the connection]
<cryptix>
btw i hope we can drop the funny random port selection by the end of the week and just use /tcp/0 everywhere
<whyrusleeping>
that would be nice to have
<ralphtheninja>
cryptix: :)
<ipfsbot>
[go-ipfs] whyrusleeping created fix/prov-store (+1 new commit): http://git.io/vkXzO
<ipfsbot>
go-ipfs/fix/prov-store 44656bd Jeromy: fix rampant memory leak in providers records storage
<zorun>
cryptix: are you confusing multipath and multicast?
nessence has joined #ipfs
<cryptix>
zorun: your right, that was a misnomer on my part
<lgierth>
cryptix: tcp/0 instead of tcp/4001?
<cryptix>
zorun: i wasnt sure if sctp also calls it multicast
nessence has quit [Remote host closed the connection]
<whyrusleeping>
lgierth: for tests, so we dont have port collisions
ei-slackbot-ipfs has quit [Remote host closed the connection]
<lgierth>
aight
<cryptix>
lgierth: we have some port selection weirdnesses in the tests. most of the system supports setting /tcp/0 and letting the kernel figure out which one is free
ei-slackbot-ipfs has joined #ipfs
<lgierth>
was already thinking how to acquire that info out-of-bind before dialing :)
<whyrusleeping>
intels throwing money around this morning
<cryptix>
lgierth: i think for swarm it already works, i have a PR in flight which will make it work for http gw and api - we just need a way to let the cli find them
<cryptix>
whyrusleeping: altera, huh? thats pretty.. well.. idk yet. even less competition in the fpga markt was not what i was hoping for
<zorun>
cryptix: in sctp, they call that "multihoming support", this is basically multipath
<zorun>
like in multipath TCP
<whyrusleeping>
cryptix: yeah, kinda weird. we will see how it pans out
<cryptix>
zorun: so its not 1 to n ? hrm
<zorun>
nope, 1 to 1, but each party can have multiple adresses and change them over time :)
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<cryptix>
zorun: yea thats also something id like to see more adoption off as well
<zorun>
well, multipath TCP is starting to gain some adoption :)
<ipfsbot>
[go-ipfs] whyrusleeping opened pull request #1315: fix rampant memory leak in providers records storage (master...fix/prov-store) http://git.io/vkXrL
<zorun>
SCTP, however, seems to be dead in practice :/
<zorun>
(mostly because of middleboxes, which is why the people behind multipath TCP were very careful to look like regular TCP)
<cryptix>
zorun: i thought it was dead too and then i saw that webrtc uses it.. weirded me out actually. i guess they chose it because they way you can configure it to be more udp-like.. idk
<zorun>
uh? didn't know that
<cryptix>
zorun: thats why i asked if sctp supports multicast :)
<zorun>
interesting
<zorun>
it seems it's used on top of UDP, though
<zorun>
which solves the deployment problem wrt firewalls & co
<zorun>
cryptix: thanks :)
<cryptix>
zorun: for throwing you into a rabbit hole? no problem :)
<zorun>
:D
<cryptix>
whyrusleeping: some feedback comments on PR#1315
mildred has quit [Ping timeout: 256 seconds]
<cryptix>
whyrusleeping: btw i think jenkins stopped picking up work
<whyrusleeping>
hrm... jenkins!
Bioblaze has quit [Remote host closed the connection]
tilgovi has joined #ipfs
<whyrusleeping>
cryptix: can you delete from a map while iterating over it?
<whyrusleeping>
i wasnt sure
Bioblaze has joined #ipfs
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<cryptix>
whyrusleeping: pretty sure you can - the iter is random but each key is only hit once, so...
<krl>
whyrusleeping: .deb, .rpm is the linux splash dl
<whyrusleeping>
krl: okay, hmm...
<krl>
but they are targeting developers, we want everyone and their grandma :)
<whyrusleeping>
yeah
<JasonWoof>
whyrusleeping: you adding protocol-specific stuff for audio/video? or a more general thing of nodes that want to send stuff to eachother in realtime?
<whyrusleeping>
JasonWoof: it will be generic to 'nodes who want to share content as its created'
<JasonWoof>
whyrusleeping: sweet
<whyrusleeping>
yeah, ipfs aims to be an amazing transport, we should make things generic when we can
<JasonWoof>
continuous streams would have many useful applicaitons
<JasonWoof>
text-based chat
<whyrusleeping>
JasonWoof: no need to use something as heavy as a live streaming framework for text based chat
<JasonWoof>
a more robust: ssh my-server tail -F /var/log/mail.log
<whyrusleeping>
but i guess that could work
<krl>
ideally we would be just to ship an executable with all the resources packed up inside
<whyrusleeping>
krl: agreed
<whyrusleeping>
you can statically compile all that.... right?
<krl>
either that, or add some magic wrap
<whyrusleeping>
yeah... we're gonna have to use some magic
<whyrusleeping>
i dont think we can statically bundle the ipfs binary into all that
<krl>
at the moment the electron-builder spits out a directory full of .sos and stuff
<whyrusleeping>
did we re-enable circleCI?
<whyrusleeping>
we need to have it build with the nofuse tag
<ipfsbot>
[go-ipfs] whyrusleeping force-pushed feat/patch from bb65900 to 97a1897: http://git.io/vkWAt
<ipfsbot>
go-ipfs/feat/patch 0082dc2 Jeromy: implement an ipfs patch command for modifying merkledag objects
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
rht_ has joined #ipfs
pfraze has joined #ipfs
<whyrusleeping>
cryptix: thoughts on that PR,
<whyrusleeping>
i want to move back to arrays
cjdmax has joined #ipfs
<cryptix>
feat/patch?
patcon_ has quit [Read error: Connection reset by peer]
patcon has joined #ipfs
<whyrusleeping>
no, fix/prov-store
<cryptix>
i'm not sure.. how much memory are we talking about?
<whyrusleeping>
well, maps need to hold enough empty entries to avoid collisions
<whyrusleeping>
which generally is a 2X bloat on the conservative side
Tv` has joined #ipfs
<cryptix>
need to change locations..
<lgierth>
krl: see you at werkstatt in a bit
<krl>
lgierth: yep!
<Bioblaze>
hay javascript gets Duplicated correct when ya duplicate a site?
<whyrusleeping>
Bioblaze: what do you mean?
<Bioblaze>
i`m not sure what its considered, importing, or uploading @.@
<Bioblaze>
not fully understanding in everything just yet.
<cryptix>
whyrusleeping: i reallt can't judge these tradeoffs without acutall usage numbers.. how many entries for different kind of nodes.. how often are these altered etc.. currentlyy i'd guess we need different classes like high-perf gateway with larger mem overhead and 'a bit slower' but slimmer for normal clients. i really donr know
<Bioblaze>
I got ipfs installed, working on how to add a site to the overall node, and yea just about it. loLz
<whyrusleeping>
yeah... i guess whats up there is good enough for now
domanic has quit [Ping timeout: 246 seconds]
<whyrusleeping>
Bioblaze: well when you 'add' things to ipfs, it makes a copy of the files for the local ipfs blockstore
notduncansmith has joined #ipfs
<whyrusleeping>
but if you add the same thing twice, it doesnt take up any extra space
<Bioblaze>
Ahh.
notduncansmith has quit [Read error: Connection reset by peer]
<whyrusleeping>
so if you add bootstrap version X.Y.Z
<whyrusleeping>
and so does someone else, youll have the same block, same hash
<whyrusleeping>
and someone else requesting it could request it from either of you
<Bioblaze>
so no use in wasting vauable resources
<Bioblaze>
I like that
lgierth has quit [Quit: Ex-Chat]
sharky has quit [Ping timeout: 265 seconds]
sharky has joined #ipfs
sprintbot has joined #ipfs
<whyrusleeping>
Hey everyone, for the sprint meeting, if you could all try and get the best internet connection possible, that would be great
<whyrusleeping>
if youre normally on wifi, finding an ethernet cable might help a bit
compleatang has joined #ipfs
compleatang has quit [Remote host closed the connection]
compleatang has joined #ipfs
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<daviddias>
whyrusleeping I armed myself with ethernet for today :)
<whyrusleeping>
daviddias: very good :)
<whyrusleeping>
every millisecond counts with audio/video between ~10 people
<daviddias>
also, closing 'things that sync' should help too
<whyrusleeping>
oh yeah, thats true
inconshreveable has joined #ipfs
<whyrusleeping>
pkill transmission
<whyrusleeping>
pkill dropbox
patcon has quit [Ping timeout: 272 seconds]
inconshreveable has quit [Remote host closed the connection]
notduncansmith has quit [Read error: Connection reset by peer]
<Confis>
Do I understand correctly that if I increase the `numProviders` value in `findProvidersDhtCmd`, I can discover 'all' peers? (i.e. it isn't that all clients using the DHT must change that value too, or some other value)
<whyrusleeping>
Confis: yeah, if you set that number to something really high like 50000000
<whyrusleeping>
it will keep searching until it exhausts the search space, or finds that many providers
<jbenet>
incominggggggg
<jbenet>
mafintosh: yeah i will be. i'm not living anywhere at the moment (crashing with friends), maybe we can airbnb a place together.
<jbenet>
daviddias: i want to make libp2p exist. it's mostly the `routing` + `networking` layers combined.
<jbenet>
ralphtheninja: re "it works locally but..." could i get a better description of the setup? the keys will be stored in other nodes, depending on the hash of the key.
<jbenet>
ralphtheninja: nice node!
<jbenet>
JasonWoof: nice! (re text-to-speech)
<jbenet>
cryptix: re talky, happy to try it, but in my experience everything sucks except hangouts. maybe can try skype? does anything let the user kill video download? (i.e. instead of me turning off video for you, you turn off your receipt of my video)
<jbenet>
cryptix: re "splitting sharness + go tests" that was me :)
<jbenet>
davidias: remind me to tell you about internet-wide multicast idea i had a long time ago. (we may be able to hack it up with cjdns)
<jbenet>
cryptix: multipath is different from multicast in that multipath is for a single process with two addresses (usually over different networks, like LAN + WAN)
<jbenet>
zorun: re "multipath TCP gaining adoption" yeah thanks to Apple actually.
<jbenet>
zorun: funny story, when i was a wee college student, i asked my networks prof why computers didnt split a UDP or TCP flow across all its interfaces, and merge them for the application with a smarter network stack (could even do better QoS when presented with fast-but-low-bw links and slow-but-high-bw links). he dismissed it as an inpractical idea that
<jbenet>
wouldn't even work well. four years later Multipath TCP wins best paper at NSDI. :D
<jbenet>
cryptix: jenkins is brain damaged. it's not _meant to_ be picking up work. chriscool will be looking into it i think.
<whyrusleeping>
whoa, havent seen one of those in a while
<whyrusleeping>
because pinning depends on pretty much every other subsystem in the codebase
<whyrusleeping>
and we needed to vendor the pinning code to make the migration code futureproof
<JasonWoof>
jbenet: awesome!
<jbenet>
not really things like logging
<whyrusleeping>
okay, i didnt vendor *all* of ipfs
<jbenet>
or bitswap
<whyrusleeping>
just things that pinning pulls in
<jbenet>
well it is though
<jbenet>
all of exchange/ is vendored, when it should not be running bitswap code -- it should use a blockstore that just uses the repo
<whyrusleeping>
it does
<whyrusleeping>
but that doesnt stop pin/pin.go from importing exchange
<jbenet>
yes but we can remove that, no?
<daviddias>
aye jbenet ! :) Nice, so that means for now, the plan is to have a impl in Node.js as well, got it. Do you happen to have the webwide multicast idea written somewhere, maybe in the random-ideas repo?
<whyrusleeping>
i was leaning away from mangling the source to get it working
<jbenet>
the tradeoff is "keep ipfs as it was which presumably should work" vs "understandable migration"
<whyrusleeping>
yeah...
<jbenet>
agree that it's more work to mangle source, but if something goes wrong it will be way harder to debug
<jbenet>
ffwd 6 months, someone asks "why did my 2-3 migration break? all my data is stuck!! help! \o/"
<whyrusleeping>
i dont think not having some unused code vendored will make debugging any easier
<whyrusleeping>
i could be wrong though
<jbenet>
pinning shouldnt require anything but a blockstore + the repo
<jbenet>
everything else is adding noise and complexity. better not to have it.
nessence has joined #ipfs
<jbenet>
code that's in the directory is code that is dauntingly threatening to be the problem.
<whyrusleeping>
hrm, maybe godeps is just terrible
<whyrusleeping>
it wasnt even being used
<Tv`>
i'd expect godeps to work per-repo not per-dir
<Tv`>
because that's what go get does
<whyrusleeping>
Tv`: thats the weird thing, it left some dirs out
<Tv`>
because you don't git clone a subdir..
<whyrusleeping>
unless i'm crazy
<jbenet>
godeps always copies all the code.
ashleyis has joined #ipfs
<cryptix>
sigh... glad that everyone misunderstood me..
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
ashleyis has quit [Client Quit]
<jbenet>
cryptix: hahaha :) <3
<jbenet>
cryptix: we really can't remain go-gettable. i prefer to give people ONE reliable way to get + build things than tell them 3 different ways that each m ay break in its own way.
<JasonWoof>
finally, people can make sites that share the same cache of jquery :)
<jbenet>
cryptix: Makefiles work. they're not a big deal. go get is nice but it's just wrong about the open source world.
<whyrusleeping>
jbenet: godeps does not copy all the code
<cryptix>
i just meant that it is still a way to get 'a version' of ipfs that might build.
<whyrusleeping>
i just tested it out by making a package the imports go-ipfs/blocks
<whyrusleeping>
ran godep save -r .
<whyrusleeping>
and it only vendored blocks and its deps
<jbenet>
cryptix: and i would rather _not_, because it's another thing i have to worry about, another UX thing that may break on people.
<jbenet>
cryptix: sometimes, less is more.
<jbenet>
whyrusleeping: interesting
<whyrusleeping>
yeah, so i have no idea what happened on the fs-repo-migrations thing
<whyrusleeping>
but i trimmed down a bunch of packages from it
<whyrusleeping>
and will push soon
<jbenet>
cryptix: also, note how horrible the "code.google.com" going down is
<jbenet>
cryptix: how much fucking code is still broken, and we have to manually go and fix.
<jbenet>
cryptix: libraries all over of course have not been updated.
<jbenet>
cryptix: and even when we PR people don't merge it quickly.
<cryptix>
jbenet: notice that i also wrote that 'go get' is totally broken for windows today
<whyrusleeping>
go is broken, lets use rust
<jbenet>
cryptix: undersood, i'm just making it clear that it's not a horrible thing to give up go get.
* whyrusleeping
grabs an 'intro to rust' book
<jbenet>
cryptix: "go get" is nice, but sadly wrong about the world outside monolithic repositories where everyone has R/W access.
<substack>
(like google)
nessence has quit [Remote host closed the connection]
<JasonWoof>
[how] does your http proxy decide what content-type to send?
<ogd>
"go" is nice, but sadly wrong about the world outside monolithic repositories where everyone has R/W access.
<jbenet>
JasonWoof: mime sniffing
<ogd>
bwahahahah
<ogd>
j/k
<jbenet>
lol
<jbenet>
ogd, you troll
* ogd
crawls back into the javascript hole
<ogd>
im the sarlacc of jabbascript
<whyrusleeping>
back under the bridge with you!
grncdr has joined #ipfs
<cryptix>
jbenet: ' it does not even accept that libraries must to vendor their deps.' hope you didnt mean that packages can import two different version of a packge because that really wont work
<whyrusleeping>
jbenet: pushed to fs-repo-migrations
<jbenet>
cryptix: why not? because go teaches people that using static + monkey patching is a good thing?
<Tv`>
jbenet: you must be using a different Go.
<JasonWoof>
jbenet: I'm all for URLs that don't have file extensions, but I think the files themselves should. I really don't like trying to explain to someone over email that a file is a plain-text file and they can open it in notepad or something
<whyrusleeping>
JasonWoof: use ipfs add -w
<whyrusleeping>
it preserves the filename
<jbenet>
Tv` not really, the problems with "importing the same libraries twice, at different versions" that everyone cites are two: (a) the type system uses the string path names for type comparison, not interfaces (a language deficiency) and (b) a cultural problem that makes it ok to use static variables across modules-- i.e. a db driver registers itself in some
<jbenet>
static map somewhere.
<cryptix>
jbenet: because pkgs with different import paths are essentially different types. even if you dont want to use gb, watch dave's talk. he does a better job at explaining it than i can
<JasonWoof>
whyrusleeping: hmmm... oh, right... I forgot that you can build sites so all/most of the dependencies (javascript, images, etc) are referenced only by their content (hash)
<cryptix>
sigh... i'm off
<Tv`>
cryptix: not even essentially
<Tv`>
import path is part of type identity
<jbenet>
cryptix: learn about how npm does things, give it a chance
<Tv`>
because of course it is, if you peek at the implementation
<whyrusleeping>
JasonWoof: thanks to the merkletree, you can have the best of both worlds, you get gateway/<hash>/path/to/content.type
<whyrusleeping>
i think rusts cargo is pretty neat
<jbenet>
cryptix: i've watched the talk AND i have already talked to dave about this
<JasonWoof>
whyrusleeping: doesn't that slow load time a little? (to have it resolve a local/pretty path, instead of linking directly to the hash)
<whyrusleeping>
JasonWoof: ah, yeah. it will slow it down a little bit
<whyrusleeping>
but in the grand scheme of things, with a fully primed node, its not bad
<JasonWoof>
It's really good that you can do it either way.
<JasonWoof>
relative paths are simple and great for testing
<JasonWoof>
People who are serious about app/site development can do fancy things to link directly to the hashed version
harlan_ has joined #ipfs
<whyrusleeping>
JasonWoof: yeah, or have everything under a single hash
mildred has joined #ipfs
<JasonWoof>
I've only recently made it so my html/css automatically change the urls for css and image files then those files change their contents
<JasonWoof>
I made it pretty easy, but there is added complexity
<JasonWoof>
for me it is worth it, so I don't have clients saying "it's still broken" because I forgot to tell them to hit refresh, or update the css url manually
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
lgierth has joined #ipfs
<ehd>
i have webpack add content hashes to all filenames and put fastly in front, only have the entrypoint HTML URLs are hash-less and not cached for long
<zorun>
jbenet: nice story :)
<zorun>
jbenet: it's actually quite hard to do right when you have very different/diverse links, so he was partly right :p
<JasonWoof>
right now (in normal lamp stack, not ipfs) when referencing images in CSS, I use a fancy Makefile rule and stylus to make my image urls like foo.png?m=<SHA1_HASH>
G-Ray has joined #ipfs
<JasonWoof>
that's great for stylus, but not so good for files that are not pre-proccessed before uploading
grncdr has quit [Ping timeout: 276 seconds]
grncdr1 has joined #ipfs
<JasonWoof>
ipfs seems to be doing well with speed
<whyrusleeping>
JasonWoof: we're getting better
<JasonWoof>
I've only played with the gateway, and probably only hit things that are cached, but still... was reading something rendered in the browser from markdown, it pulled 5-ish javascript files and two css files, and still loaded in a reasonable amount of time
<whyrusleeping>
resolution time can be a pain right now, but once transfers start, it should be decent
<JasonWoof>
sure, DHT is not inherrantly fast
<JasonWoof>
love that you guys care about speed
warner has quit [Quit: ERC (IRC client for Emacs 24.5.1)]
<whyrusleeping>
of course!
<whyrusleeping>
we want people to use it, and people wont use it if its not fast
<cryptix>
lgierth: interesting.. thx. i thought about interfaceing with the existing daemon but this gives more interesting options
G-Ray has quit [Quit: Konversation terminated!]
<lgierth>
the current state is that kestrel can peer with a cjdns node, but not much more
alexandria-devon has quit [Quit: alexandria-devon]
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
williamcotton has joined #ipfs
lgierth_ has joined #ipfs
lgierth has quit [Ping timeout: 272 seconds]
lgierth_ is now known as lgierth
cjdmax has quit [Remote host closed the connection]
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
vijayee__ has quit [Ping timeout: 256 seconds]
vijayee_ has quit [Ping timeout: 264 seconds]
inconshreveable has quit [Ping timeout: 265 seconds]
inconshreveable has joined #ipfs
mildred has joined #ipfs
<cryptix>
sprint in 5?
<whyrusleeping>
yeap
<lgierth>
+
<ipfsbot>
[go-ipfs] lgierth opened pull request #1316: WIP: discover peers from cjdns (master...discovery-cjdns) http://git.io/vkMG8
<harlan_>
My friend Duke is possibly interested in contributing (self booting protocol stacks / VM stuff) -- is it appropriate for him to come by sprint meeting and introduce himself?
<jbenet>
harlan_ yeah totally, we usually do a bit at the beginning about everyone meeting each other.
<jbenet>
harlan_ its settling into 10min for people sync, 15 min for updates on last week, and then 20min for next week.
<jbenet>
(roughly)
<harlan_>
gr8
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<jbenet>
hey everyone-- sorry for those locked out of the room
<jbenet>
we can try out a different thing for next week, like talky.io or something.
<jbenet>
and since these are getting very large, it may make sense to split them into a community hangout (more the first part, of people getting to know each other and high level discussion), and then smaller sync-ups between people working together on sub-thigns (like go-ipfs or node-ipfs, etc)
<jbenet>
(lmk if you have ideas on how to make these discussions more useful, funner, etc).
<jbenet>
---- ok next thing, if you know what to work on this week, just modify the etherpad as usual.
<jbenet>
if you don't, sync up with me and whyrusleeping to pick the set of things to work on
<jbenet>
(eg kbala_ lgierth, davididas most likely need to sync with me)
<jbenet>
the etherpad will be made into an issue in ~5hrs, so ok to use this time now to figure out the things to work on.
<lgierth>
https://palava.tv/ is another option, not sure how well it works with > 3 people
<whyrusleeping>
lets just write an ipfs video chat app
<whyrusleeping>
p2p! :D
<kbala_>
jbenet hey, what should i aim to do this week?
<lgierth>
the shed should be purple!!
<lgierth>
:)
<jbenet>
kbala_ i think the thing to focus on this week will be formalizing the model for bitswap -- both protocol and decision making
<lgierth>
jbenet: let me know about the gateway stuff, i'll be at the meetup another hour or so, then head home, then be available until ~ 1 am utc
<jbenet>
so that we can figure out how ML driven decisions will be fed into the protocol
<jbenet>
and how can extract + use features
<jbenet>
lgierth: sounds good-- i'll make a repo for gateway things and write out issues in the next hour or so
<lgierth>
ack
<jbenet>
kbala_ maybe we can meet 1-to-1 to go over bitswap first, and then go over a plan to do this + make a tasklist ?
<kbala_>
jbenet sounds good! when would you like to meet?
rht_ has quit [Quit: Connection closed for inactivity]
<jbenet>
kbala_ let's discuss 13:30-14:00 PST or 14:00-14:30 ?
<jbenet>
daviddias: also if you want to sync up re node-ipfs and best direction, let's do that. would be good to include krl tperson and mappum as they're also on js stuff
<harlan_>
+1 to community hangouts. not sure of a good free solution for >10 people, but I've had great experience with http://zoom.us , which for $10/mo will give you 25 participants, which is probably as large as you want a group to be anyway...
shurcooL has quit [Ping timeout: 255 seconds]
<kbala_>
jbenet 14:00-14:30 is good
<lgierth>
did my microphone work ok?
<jbenet>
lgierth: yeah it did!
<lgierth>
it was actually the first time i ever got it to work
<whyrusleeping>
lgierth: yep!
<lgierth>
cool
<jbenet>
harlan_ yeah, thanks! will check it out. know if it works on linux?
notduncansmith has joined #ipfs
<daviddias>
jbenet: lets go for that, was just preparing a issue on node-ipfs for basic spec compliace to make the thing work :)
notduncansmith has quit [Read error: Connection reset by peer]
<whyrusleeping>
i dont bother trying to video chat on linux
<whyrusleeping>
apple doesnt release their video drivers
<jbenet>
update for people not able to get on the call: this week is fixing a lot of random things too. biggest single efforts of "new thigns" are getting a node-ipfs up to talk to go-ipfs, and electron app for users who dont want to run a daemon from cli
<jbenet>
wking, rht, chriscool, gatesvp: drop an update here when you get a chance + edit etherpad as needed
<daviddias>
jbenet let's do at 14:30h, so I can finish organising the ideas a issue first :)
<whyrusleeping>
jbenet: its accurate afaik :/
<jbenet>
daviddias: +1
<ipfsbot>
[go-ipfs] whyrusleeping pushed 1 new commit to fix/prov-store: http://git.io/vkMPA
<ipfsbot>
go-ipfs/fix/prov-store d796377 Jeromy: use map and array combo for better perf
<cryptix>
jbenet: hrmpf.. there is still a race in the progressbar thing..
<whyrusleeping>
cryptix: are we using it wrong?
<kyledrake>
jbenet whyrusleeping So, this morning I think I may have tracked the issue. I accidentally forgot to remove a synchronous archiver call in our code, so it was duplicating a lot of import tasks (and making our uploads go a lot slower), and running them sync without a throttle, I'm pretty sure this was -the- problem.
<whyrusleeping>
kyledrake: well that makes me happier :P
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<kyledrake>
I haven't restarted ipfs all night (turned off the hourly reboot) and it's been smooth sailing so far now that I fixed that.
<whyrusleeping>
hell yeah!
<jbenet>
yay \o/
<whyrusleeping>
i want to have a virtual office. where we all have vr headsets and are in the same virtual room
<cryptix>
whyrusleeping: could be.. need to take a closer look
alexandria-devon has joined #ipfs
<jbenet>
whyrusleeping: that's a sweet idea. me too
<daviddias>
whyrusleeping that would be rad! Not just for conf calls, but imagine, code everywhere (forest, beach, park, etc) without any sun glare and with any screen size you want, all you need is the VR googles and a keyboard
<kyledrake>
whyrusleeping the replica ipfs add -r issue isn't related, so I just started a command that will allow me to get that stack dump for you
<cryptix>
whyrusleeping: you might be right - these locations are different then before
Bioblaze has quit [Remote host closed the connection]
PayasR has joined #ipfs
harlan__ has joined #ipfs
<harlan_>
?
harlan_ has quit [Quit: Page closed]
<daviddias>
jbenet just realized that when you said "now-14:00h, or after 14:30h" it was actually a question with 3 options and not 2, right now is 13:30h on your clock, right? Let's do it at 14:00h then :) //cc krl tperson mappum
<jbenet>
daviddias: oh i meant two options, (now -14:00) or (14:30, --) becasue am chatting with kbala 14:00-14:30
<wking>
jbenet: Etherpad edited. I'll need decisions on go-ipfs#1291 before we can land the tail commits from go-ipfs#1136.
<jbenet>
wking: ok thanks!
<harlan__>
btw jbenet zoom.us is what we used when I interviewed you, so you might have linux experience on it...
<daviddias>
got it, 14:30h still works fine for me then :)
notduncansmith has quit [Read error: Connection reset by peer]
<whyrusleeping>
huh... theres no longer a Resolve method on the core.IpfsNode
<whyrusleeping>
i missed something
<jbenet>
wking: right. i think a fork is likely the better option for now. want me to weigh in there?
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
lgierth has joined #ipfs
PayasR has quit [Remote host closed the connection]
<kbala_>
jbenet want to talk here or on hangouts?
<wking>
jbenet: fork sounds like the current suggestion there, so probably no point in commenting with that. If you have something else to add, fire away though ;)
<jbenet>
kbala_ hangouts may be higher throughput for this-- let's do that. maybe take notes and we can add them to https://github.com/ipfs/notes/
<JasonWoof>
fwiw, on x86 linux here, I've run zoom.us inside virtualbox (with one of the free windows images from modern.ie) and it worked well
<JasonWoof>
I don't have a camera, so I can't confirm if that part works, but I could hear and see just fine, and they could here me. Just had to remember to set my headset as the default microphone in pavucontrol
ashleyis has joined #ipfs
ashleyis has quit [Client Quit]
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
kbala_ has quit [Ping timeout: 276 seconds]
jonl has quit [Ping timeout: 276 seconds]
kbala_ has joined #ipfs
jonl has joined #ipfs
Bioblaze has joined #ipfs
inconshr_ has quit [Read error: Connection reset by peer]
<ipfsbot>
[go-ipfs] whyrusleeping created builder-default (+1 new commit): http://git.io/vkDk5
<ipfsbot>
go-ipfs/builder-default ba18dac Jeromy: make the default repo for corebuilder work
<ipfsbot>
[go-ipfs] whyrusleeping opened pull request #1317: make the default repo for corebuilder work (master...builder-default) http://git.io/vkDIE
<JasonWoof>
wow, a full implementation in node?! this must mean that you aren't going to have your protocols be defined by what your illegible java sources do?
<lgierth>
java sources?
<JasonWoof>
sorry, still healing from the trauma of freenet
<lgierth>
heh
<lgierth>
i was going to ask *exactly* that
<lgierth>
:P
<JasonWoof>
I think the final straw for me was when I complained about the high memory consumption, despite my conservative settings, and they told me that it didn't run well with less than some huge amount of memory (which was about 3X what my machine had in actual RAM)
guest449 has joined #ipfs
<JasonWoof>
oookay, clearly that thing is never going to work well
<lgierth>
i'm kinda spoiled by cjdns
<lgierth>
it has a hard memory cap of 8 MB
<Evermore>
JasonWoof: I get that from a lot of people these days
<Evermore>
I decided not to run bitcoin since it was going to need more RAM than I had in my entire house for the database
<Evermore>
Between that and the CPU churning to catch up to the log it was too boring
<JasonWoof>
cjdns has been great. I'm not on hyperboria, just use it as a vpn between my machines.
<JasonWoof>
bitcoin is weird
<mmuller_>
it's leaner than it used to be.
<JasonWoof>
I have some trouble with the "ahh, whatever, we'll just have everybody store the entire history of every financial transaction ever" attitude
<JasonWoof>
I realize that not all clients have to
<JasonWoof>
but the one I tried did
<mmuller_>
I guess I'm running it on a 2G machine
<jbenet>
kbala_ what's your github username/
<Evermore>
I think someone has to do that, but there should be a way to break off a trusted history at some point
<Evermore>
I don't want to dedicate 30 GB to that log when everyone agrees what happened
<kbala_>
jbenet heems
<whyrusleeping>
JasonWoof: i agree re: bitcoin
<Evermore>
JasonWoof: I also had someone tell me it was normal that my Lua program was leaking 4 bytes per frame because garbage collection is not mandatory or something.
<jbenet>
daviddias: ok done chatting with kbala_ ready to chat about node-ipfs whenever. (no rush though)
<harlan__>
Evermore: my friend John Light is part of the https://bitseed.org/ project -- bitcoin server on a $150 plug in server...
<Evermore>
harlan__: nice
<JasonWoof>
I like the method of buliding a reference implementation and experimenting with a language that is fun and easy, like coffeescript+node or go or whatever. Then later, once you've worked out the protocols/etc, implement it in C
<Evermore>
Course for $300 I could probably get a little destop that could run bitcoin and 3 other big programs
<mappum>
jbenet: i'm here to talk about node-ipfs/cube
<daviddias>
jbenet: I'm ready too :) - beta.talky.io/node-ipfs
<harlan__>
sure
notduncansmith has joined #ipfs
<jbenet>
cool let's jump on that call
<daviddias>
(let's try this time talky)
notduncansmith has quit [Read error: Connection reset by peer]
<jbenet>
cc krl tperson if you're interested o/
<JasonWoof>
lgierth: I hope to do that one day. Currently I'm giving the guy downstairs $20/month to use his wifi
<jbenet>
mappum o/
<JasonWoof>
lgierth: but cjdns is great at getting my laptop and desktop to talk to eachother. I'm pretty sure it doesn't bounce the signal off my server when I've got both machines at home
guest449 has quit [Remote host closed the connection]
<JasonWoof>
openvpn on my server gets them talking to eachother, but I want to be able to do large file transfers while at home without the data going to my server and back. cjdns can automatically switch between direct sending and bouncing off server
<lgierth>
JasonWoof: if the laptop and desktop are peered, yeah
<JasonWoof>
lgierth: yeah, desktop initiates connection to server. laptop initiates connection to server and desktop
<lgierth>
they can peer automatically if they're on the same network
<lgierth>
ETHInterface
mildred has quit [Quit: Leaving.]
<JasonWoof>
seemed like that would make it automatically peer with anybody over eth
harlan__ is now known as harlan_
<lgierth>
yep
<tperson>
jbenet: you dropped out
<jbenet>
yeah interesting
<jbenet>
i refreshed and i cant get back in
<jbenet>
mappum any luck?
<whyrusleeping>
jbenet: is it safe to assume that if a nodes ipns entry is not in the local datastore, they dont have one?
<JasonWoof>
lgierth: the desktop has a stable local IP address, so I just told the laptop to connect to that
foobar___ has joined #ipfs
chriscool has joined #ipfs
therealplato has quit [Ping timeout: 252 seconds]
foobar___ has quit [Client Quit]
foobar___ has joined #ipfs
<lgierth>
JasonWoof: yeah that's fine -- with ethinterface you'll also get a tiny bit more payload per underlying ethernet frame since the cjdns packets aren't wrapped in a UDP datagram. but that's negligible
Wallacoloo has joined #ipfs
foobar___ has quit [Client Quit]
guest449 has joined #ipfs
alexandria-devon has quit [Quit: alexandria-devon]
lidel has left #ipfs [#ipfs]
lidel has joined #ipfs
<JasonWoof>
oh :) I didn't realize it didn't use UDP
<JasonWoof>
that's cool
<JasonWoof>
I think I set my laptop to also try to connect to my home computer via openvpn
<JasonWoof>
cjdns can't connect some places, I'm guessing because they block udp
guest965 has joined #ipfs
<JasonWoof>
lame to have that many layers, but... worth it for the convenience of having the cjdns ipv6 addresses just keep working no matter what
<lgierth>
try binding it to port 80 :P
therealplato has joined #ipfs
<lgierth>
it unfortunately doesn't make a lot of sense to use tcp as a transport
<lgierth>
you'd end up with tcp-over-tcp
inconshreveable has joined #ipfs
<lgierth>
which is a recipe for disaster, just like tcp over 3g/4g, which has its own retransmission mechanism
chriscool has quit [Ping timeout: 244 seconds]
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
<ralphtheninja>
jbenet: so I tried this .. ipfs dht put foo bar .. on one machine .. tried ipfs get foo on another machine and got an error
<ralphtheninja>
jbenet: but when I switched machines and did put on the other machine and get on the other, it worked
<ralphtheninja>
jbenet: might be some network problems at work or something like that
<whyrusleeping>
ralphtheninja: what error did you get?
<ralphtheninja>
'error: dial attempt failed: peer has no address'
<JasonWoof>
I just read "Making your own ipfs service"... I'd like to make one where my friends and I can tell eachother to pin things
<JasonWoof>
would a custom service have access to do that? if so, can I say "download and pin this" or do I have to wait for it to be downloaded to pin it?
ebarch has quit [Ping timeout: 256 seconds]
<JasonWoof>
at first I was thinking about backups (probably not a great way to do it), but then I was thinking this would be very cool for stuff I publish
<JasonWoof>
would be very cool if a few friends would automatically download/pin stuff I publish, so my stuff stays up when/if I go offline
<JasonWoof>
or... maybe it would be useful for encrypted, personal backups.
<JasonWoof>
nobody's going to request/cache my backups, but my friends would if I asked them to
guest965 has quit [Ping timeout: 256 seconds]
<whyrusleeping>
JasonWoof: a custom service would be able to do that, yes
<jbenet>
kbala_ (there's probably other papers to look at, but that's a good one to start with)
<jbenet>
whyrusleeping: "is it safe to assume that if a nodes ipns entry is not in the local datastore, they dont have one?" no, because crashes and multiple writers.
<jbenet>
whyrusleeping: routing may have one
<whyrusleeping>
jbenet: okay, that complicates daemon startup...
<jbenet>
ralphtheninja: not sure if that should work that way. whyrusleeping is your best bet -- he wrote all that
<jbenet>
whyrusleeping: hm. that makes sense. this would be simplified if we make the IPNS entry corresponding to the node's ID be managed by the local node always (i.e. its a representation of local node state)
<jbenet>
whyrusleeping: cause in that case we can make that assumption
<jbenet>
whyrusleeping (or overwrite what's out there)
<jbenet>
whyrusleeping (safely)
<jbenet>
whyrusleeping -- but for names in general, we shouldn't assume that.
<jbenet>
JasonWoof " I'd like to make one where my friends and I can tell eachother to pin things" we want this too! help us bake it into go-ipfs. maybe can start a discussion on how to make this over at github.com/ipfs/notes
guest9651 has joined #ipfs
guest965 has quit [Read error: Connection reset by peer]
guest9651 has quit [Read error: Connection reset by peer]
guest965 has joined #ipfs
guest965 has quit [Read error: No route to host]
guest965 has joined #ipfs
tilgovi has quit [Ping timeout: 264 seconds]
<kyledrake>
Are you currently using S/Kademlia for the DHT?
<kyledrake>
Or is that a down-the-road thing
<whyrusleeping>
kyledrake: we are just using kademlia
<kyledrake>
cool.
<jbenet>
kyledrake: more down the road.
notduncansmith has joined #ipfs
notduncansmith has quit [Read error: Connection reset by peer]
lgierth_ has joined #ipfs
Guest77051 is now known as amiller_
amiller_ has joined #ipfs
amiller_ is now known as amiller
lgierth has quit [Ping timeout: 272 seconds]
tilgovi has joined #ipfs
neoteo has quit [Quit: Be back later ...]
<ipfsbot>
[go-ipfs] whyrusleeping created feat/keypkg (+1 new commit): http://git.io/vkDKf
<ipfsbot>
go-ipfs/feat/keypkg ef29443 Jeromy: move util.Key into its own package under blocks
ashleyis has joined #ipfs
<ipfsbot>
[go-ipfs] whyrusleeping opened pull request #1318: move util.Key into its own package under blocks (master...feat/keypkg) http://git.io/vkDKO