IlanGodik has quit [Quit: Connection closed for inactivity]
<achin>
i need some git help (maybe cc richardlitt or cc whyrusleeping).
nonaTure has quit [Quit: Leaving.]
Encrypt has quit [Quit: Sleeping time!]
gperrin has joined #ipfs
<achin>
i'm trying to figure out why commit 4a33683443703f3ef06c8ec3d1dc7bcffb8a792d in go-ipfs has an AuthorDate of May20,2016, but a CommitDate of Jan12,2016
gperrin has quit [Client Quit]
dignifiedquire has quit [Quit: Connection closed for inactivity]
<achin>
ooooh
<achin>
i think there is some funkiness due to all the rebasing that happened during the dev040 merge into master
<achin>
this will require some pondering
tmg has joined #ipfs
neurosis12 has joined #ipfs
neurosis12 has quit [Changing host]
neurosis12 has joined #ipfs
grahamperrin has left #ipfs ["Leaving"]
gperrin has joined #ipfs
gperrin has quit [Client Quit]
grahamperrin has joined #ipfs
grahamperrin has left #ipfs ["Leaving"]
fe has joined #ipfs
fe is now known as Guest90035
fiatjaf_ has quit [Ping timeout: 240 seconds]
<Qwertie>
Is there a systemd config or something I can use to have the ipfs daemon run all the time?
Guest90035 has quit [Remote host closed the connection]
<ipfsbot>
go-ipfs/master bc94293 rht: Merge pull request #2223 from ipfs/feature/doc-add...
<whyrusleeping>
vakla: no
<vakla>
:(
treaki has quit [Ping timeout: 240 seconds]
<ipfsbot>
[go-ipfs] rht closed pull request #2228: Fixed up docs for bitswap (master...docs/bitswap) https://git.io/vzux0
hellertime has joined #ipfs
<ipfsbot>
[go-ipfs] rht closed pull request #1826: Update help text to avoid confusion of where files can be added from (dev0.4.0...patch-2) https://git.io/vCnI2
elima_ has joined #ipfs
<ipfsbot>
[go-ipfs] rht opened pull request #2233: Change all mfs references to unixfs (master...patch-2) https://git.io/vzwS8
fe has joined #ipfs
fe is now known as Guest24384
e-lima has quit [Ping timeout: 272 seconds]
linton_s_dawson has joined #ipfs
Guest24384 has quit [Ping timeout: 264 seconds]
linton_s_dawson has quit [Quit: Leaving.]
voxelot has quit [Ping timeout: 265 seconds]
gaboose_ has quit [Quit: Connection closed for inactivity]
<ipfsbot>
[go-ipfs] whyrusleeping closed pull request #2134: add helper command for publishing out of files api (dev0.4.0...feat/files-publish) https://git.io/vESM4
<ipfsbot>
[go-ipfs] whyrusleeping created feat/files-publish (+2 new commits): https://git.io/vzwHg
<ipfsbot>
go-ipfs/feat/files-publish e0f4cb3 Jeromy: add helper command for publishing out of files api...
<ipfsbot>
go-ipfs/feat/files-publish 75f267c Jeromy: fix 'local' publish/resolve for testing...
<aaaaaaaaa____>
hi - we have a large archive and some people who want to make backups of it. we are currently adding everything into ipfs, but i'm unsure about the "proper" way to proceed from there. for example, would we "ipfs pin ls --type=recursive > hashes.txt" and then make that available for a shell script on the backup side to pin every hash in the list?
<aaaaaaaaa____>
or is there a smarter way to do it? still kind of picking it all up
The_8472 has quit [Ping timeout: 240 seconds]
auspiciously has quit [Read error: Connection reset by peer]
nonaTure has joined #ipfs
turtosa has joined #ipfs
jfis has joined #ipfs
fe has joined #ipfs
Tanizaki2 has joined #ipfs
fe is now known as Guest55976
IlanGodik has joined #ipfs
pfraze has quit [Remote host closed the connection]
tanizaki has quit [Ping timeout: 250 seconds]
<jbenet>
aaaaaaaaa____ yep that will work
voxelot has quit [Ping timeout: 260 seconds]
<jbenet>
aaaaaaaaa____ i'd make sure to call `ipfs refs $hash` first-- which gives more feedback to a pin operation
tmg has quit [Ping timeout: 272 seconds]
felixn_ has joined #ipfs
<whyrusleeping>
chriscool: it calls pinner.IndirectKey()
<whyrusleeping>
which as far as i can tell doesnt actually exist...
Tanizaki2 has quit [Ping timeout: 260 seconds]
felixn has quit [Ping timeout: 265 seconds]
tanizaki has joined #ipfs
tanizaki has quit [Read error: Connection reset by peer]
The_8472 has joined #ipfs
tmg has joined #ipfs
ugjka has quit [Ping timeout: 272 seconds]
<chriscool>
whyrusleeping: it's IndirectKeys(), with an "s" and it exists line 292 in ipfs-2-to-3/Godeps/_workspace/src/github.com/ipfs/go-ipfs/pin/pin.go
<whyrusleeping>
right
<whyrusleeping>
but it doesnt exist in the current codebase
<chriscool>
maybe we have to debug the go-ipfs that is vendored in fs-repo-migrations then
<whyrusleeping>
yeap, i'm working on it
<jbenet>
it may be useful to vendor it with the git sha in the go-ipfs repo. (i.e. name the directory that, or a file in the dir)
<chriscool>
I wonder what will happen when maybe migration 3 to 4 will need a newer version of go-ipfs
<chriscool>
we will have to vendor many different go-ipfs versions in fs-repo-migrations
computerfreak has quit [Quit: Leaving.]
<jbenet>
chriscool we can vendor them with gx possibly.
adamc1999 has quit [Ping timeout: 260 seconds]
<chriscool>
ok
<jbenet>
chriscool and if not, it's not a big deal. code is small. git deduplicates in transit.
<chriscool>
yeah ok
<jbenet>
what we should definitely do is make sure the binary doesn't explore
<jbenet>
each migration should be it's own binary, and fs-repo-migrations should be a downloader/runner for them.
<jbenet>
(but it's not critical yet, that improvement can come later)
<jbenet>
maybe we can say when fs-repo-migrations bin size hits 30MB
<whyrusleeping>
^
<whyrusleeping>
yes
<chriscool>
yeah in fact it looks like go-ipfs is vendored inside ipfs-2-to-3/ not directly under fs-repo-migrations
<whyrusleeping>
yep
<whyrusleeping>
its its own version
ygrek has quit [Ping timeout: 265 seconds]
<jbenet>
right. its so that we can split that out into another bin easily
<whyrusleeping>
jbenet: question you might have an answer to
<whyrusleeping>
'ipfs files cp'
<whyrusleeping>
lets you copy any ipfs hash into the files api namespace
<whyrusleeping>
but
<whyrusleeping>
it does so just as a link
<whyrusleeping>
it doesnt fetch that entire file
<whyrusleeping>
or dag
<whyrusleeping>
this is an issue because the files api root is pinned
felixn_ has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
nonaTure has quit [Quit: Leaving.]
<whyrusleeping>
i *do* like the UX of having 'ipfs files cp' stay a constant time operation (and not fetching anything)
ugjka has joined #ipfs
<whyrusleeping>
chriscool: commit a70e0ecca8b457f6ce1687202e360e38491c854b fixes the issue
<whyrusleeping>
i pushed it to the repo (not master)
<chriscool>
whyrusleeping: great, thanks I will rebase on top of it
<jbenet>
whyrusleeping: well the files api root should not be pinned. i've mentioned this a few times on github. we can have a `/pins` thing that behaves like what you want `files` to behave like.
<jbenet>
so "$reporoot/files" is just files. "$reporoot/pins" is all the pinned stuff available fs/path style.
<jbenet>
I have lots of files that i want to be able to address nicely and beautifully with the files api that i DO NOT want to download, for they are huge.
<ipfsbot>
[go-ipfs] RichardLitt force-pushed feature/squash-note from 7711fce to 1b97a0f: https://git.io/vzwpu
<ipfsbot>
go-ipfs/feature/squash-note 1b97a0f Richard Littauer: Added squash note to contribute.md...
<whyrusleeping>
jbenet: hrm... thats difficult then
<whyrusleeping>
i also don't remember you ever saying that the files root shouldnt be pinned
<whyrusleeping>
i remember you saying that we shouldnt replace pinning with that
<aaaaaaaaa____>
jbenet: thanks! Will test out soon
<whyrusleeping>
chriscool: you might just want to cherry pick it
<whyrusleeping>
its on top of your code
<whyrusleeping>
(i just couldnt push to your branch)
<chriscool>
whyrusleeping: ok I will cherry-pick it, thanks!
<jbenet>
whyrusleeping: why is it difficult?
<whyrusleeping>
well, things like the npm-registry want to stay pinned
<whyrusleeping>
but they also change constantly
<whyrusleeping>
and repo gc is *completely* unsafe if the user is doing anything with the files api
<whyrusleeping>
but
<whyrusleeping>
if we set that precedent, then it makes the other stuff i was doing easier
<whyrusleeping>
since i don't have to worry about pin safety
<jbenet>
whyrusleeping running repo gc being completely unsafe at any point is a bug, and an important one to fix. i should be able to run a cron job every minute with ipfs repo gc back to back and be _safe_, even if that's super inefficient.
<whyrusleeping>
then we need to pin the files api root?
<jbenet>
i don't understand the tension
<whyrusleeping>
echo "this is some random stuff" | ipfs files write /a/b/c
<whyrusleeping>
ipfs repo gc
<jbenet>
ipfs files api is jut files api, it's not pinning.
<whyrusleeping>
ipfs files read /a/b/c
<jbenet>
pinning is a separe thing
<jbenet>
yeah /a/b/c should no longer be in your repo
<jbenet>
you can address it, just like you can address any other object.
<jbenet>
it may not be in your repo, and will try to find it in the network
<whyrusleeping>
and wait forever
<jbenet>
not with timeout flag
<jbenet>
or if the content exists
<whyrusleeping>
so its not unsafe, it just breaks a lot of things
<jbenet>
no, it IS safe. safe is safety of content according to the api we expect.
<whyrusleeping>
think of the npm-registry usecase
<jbenet>
what you mean is `ipfs files write --pin=true /a/b/c`
<jbenet>
which should add it to the files root, and then pin `/a/b/c` specifically.
<whyrusleeping>
and what happens on a second call to 'ipfs files write --pin-true /a/b/c'
<whyrusleeping>
is the previous node unpinned?
<whyrusleeping>
what if its pinned elsewhere for some other reason?
<jbenet>
good question. im not sure, it could be another flag, like --pin-replace.
<jbenet>
like, i get the use case you want
<jbenet>
and it's a good one for sure
<jbenet>
and maybe we should have exactly that
<jbenet>
but we should also have the files api without forcing pin.
<jbenet>
the files api is awesome and i want to use it with huge files in the network, but not necessarily pinned in the node i'm adding them to
<whyrusleeping>
we should probably just drop all the pin locking
<whyrusleeping>
if thats the case
<whyrusleeping>
it simplifies a lot of things
<jbenet>
whyrusleeping: we cant drop pin locking i dont think, it's needed to halt a repo gc while an add happens.
<jbenet>
not halt, prevent*
<whyrusleeping>
not all pin locking, i just meant the stuff in the files api code
<jbenet>
oh.
<whyrusleeping>
should have clarified
<jbenet>
i mean, we can do that if you're comfortable with removing that use case. i do agree with you that it's very nice not to have to think about it at all.
<jbenet>
like pins can be annoying. its important to have as a plumbing thing for full control, but i do like the porcelain of having the tool in a mode that does all the right things you'd expect in the normal case.
<chriscool>
could you take a look at merging the PR?
<whyrusleeping>
jbenet: yeah... I agree with you
<whyrusleeping>
we just need to REALLY make sure people know about that
<whyrusleeping>
i can see people relying on it for things
<whyrusleeping>
running a gc, expecting everything to be fine
<whyrusleeping>
and losing tons of work
disgusting_wall has quit [Quit: Connection closed for inactivity]
<whyrusleeping>
(like daviddias did)
<whyrusleeping>
(for this exact reason)
<jbenet>
yep, i understand.
<whyrusleeping>
maybe
<whyrusleeping>
ipfs files pin /a/b/c
<whyrusleeping>
would change the pin every time its flushed past there?
fe has joined #ipfs
<jbenet>
and it would even make sense for `ipfs files {write, cp}` to have --pin=true as a default, to match `ipfs add`
<jbenet>
whyrusleeping: oh i like that
fe is now known as Guest59742
<daviddias>
you mean, when the process crashed because the server ran out of space?
<jbenet>
whyrusleeping: yeah "ipfs files pin /a/b/c" is asking to pin _that path_, not the file pointed to.
<whyrusleeping>
daviddias: yeah, and we ran a gc?
<whyrusleeping>
jbenet: yeap
<jbenet>
whyrusleeping: i like that. should be very clearly documented in `ipfs files pin` then.
<jbenet>
whyrusleeping: so to get what you want, `ipfs files pin /`
<jbenet>
an there you go?
<whyrusleeping>
yeap!
<whyrusleeping>
sounds about right to me
<daviddias>
whyrusleeping: yeah, tough times
<jbenet>
would this path show up in the `ipfs pin ls` ?
Guest55976 has quit [Ping timeout: 264 seconds]
<jbenet>
would this be a different type of pin? or it would just show the partial path, like "files/a/b/c"
<whyrusleeping>
i think ipfs pin ls would show the latest hash
<jbenet>
how do we differentiate them? i feel like the pin entry should have the path so can reason about it?
<whyrusleeping>
'ipfs files pin ls'
<whyrusleeping>
would show paths
adamc1999 has joined #ipfs
<whyrusleeping>
'ipfs pin ls' would show the latest bubbled up hash
<whyrusleeping>
and be just as confusing as everything else currently using pinning for things
<jbenet>
whyrusleeping ok sounds good.
<jbenet>
hahahahahha
<jbenet>
:D
<jbenet>
yaaay candy mountain
<whyrusleeping>
lol
<daviddias>
idea: have pinning associated with the type of data, e.g: ipfs block pin <hash> (pin just a block); ipfs object pin <hash> (pin an object and also offers ipfs object pin -r to traverse the dag and pin everything); ipfs files pin <path> (pins that path, pinning the actual state and future mutations)
<jbenet>
daviddias: nice yeah that goes along with "its a different type" above.
<jbenet>
this is relatively easy to implement, one more pin set whose semantics are different.
<ansuz>
Kubuxu: I'm thinking of migrating from linode to OVH soon
<ansuz>
seems like linode gets DDoS'd every week lately
<Kubuxu>
ansuz: VPSes in OVH are real lottery, I had one that was getting rebooted few times a month, my friend had (there was kernel update recently) one with 3 months up-time.
raxrb has joined #ipfs
Samuel has joined #ipfs
<Kubuxu>
OVH Dedis great, soyoustart is still ok, kimsufi a little crappy.
<lidel>
Kubuxu, we could add "Copy web+fs: address" as an additional option, but perhaps a better way to approach this is to add "Canonical address prefix" property to the Preferences screen. That way one person could decide to leave it empty (default) and would copy /ipfs/.. , someone else want web+fs:, someone else could even put an address of own http gateway.
gaboose has joined #ipfs
<Kubuxu>
If we get The_.8472 to finish the protocol handling it will be unnecessary.
reit has joined #ipfs
<Kubuxu>
lidel: ^^
<lidel>
Yeah, then you could simply copy from the address bar :)
<The_8472>
hrrm, i guess I should try that separately. i was trying to do too many things at once. supporting fs: and canonizing http: -> fs:
<ipfsbot>
[js-ipfs-api] Dignifiedquire force-pushed chore/updates from 97f399f to d9039de: https://git.io/vzrWz
<ipfsbot>
js-ipfs-api/chore/updates d9039de dignifiedquire: test: Remove setup api config not needed in 0.3.11 anymore
Quiark has quit [Ping timeout: 245 seconds]
adamc1999 has joined #ipfs
pfraze has quit [Remote host closed the connection]
<patagonicus>
Huh? The docker image marked :latest is back to 0.3.11. Wasn't that a 0.4 build before (a repo I've created with one of the docker images has the newer format, so either :latest or :release used to be 0.4)?
pfraze has joined #ipfs
<lidel>
jbenet, shouldn't :release be 0.3.x and :latest 0.4.x?
adamc1999 has quit [Ping timeout: 250 seconds]
<ipfsbot>
[js-ipfs] diasdavid force-pushed feat/blocks from 128f946 to 8836e17: https://git.io/vzrlz
<ipfsbot>
js-ipfs/feat/blocks 245918f David Dias: core: block get + test
<ipfsbot>
js-ipfs/feat/blocks 8836e17 David Dias: move block get to use BlockService and Block data struct
<mafintosh>
whyrusleeping: what was that network + docker testing thing we talked about called?
<whyrusleeping>
mafintosh: i was actually working on writing a nicer more configurable thing
<whyrusleeping>
that one just has some tooling around packet loss, bandwidth and latency
<mafintosh>
right
<whyrusleeping>
i'm working on sketching up a format for a way to define different subnetworks and do NAT stuff
<mafintosh>
one for testing hole punching envs would be great
<mafintosh>
yea perfect
hashcore has quit [Quit: Leaving]
jaboja has joined #ipfs
<Kubuxu>
Hmm, I am thinking if it is possible to create hash shortening service, something along frak security I just want short link.
<Kubuxu>
Standardized WoT on IPFS would make a lot of things much easier.
chriscool has quit [Quit: Leaving.]
chriscool has joined #ipfs
reit has quit [Ping timeout: 264 seconds]
Encrypt has joined #ipfs
pfraze has quit [Remote host closed the connection]
<Codebird>
I am preparing to IPFS my public domain music site. Should be fun.
<ipfsbot>
[js-ipfs-api] diasdavid deleted chore/updates at d9039de: https://git.io/vzr4z
pfraze has joined #ipfs
<Codebird>
Step one: rsync my current collection up to the remote server. Step two: Create indexs on remote server. Step three: Rsync it back down again. Step four: IPFS. Step five: Go enjoy free music.
reit has joined #ipfs
computerfreak has quit [Quit: Leaving.]
xiaomorph is now known as yangwao
pfraze has quit [Remote host closed the connection]
<Kubuxu>
Is there a way to check from API if given object is available locally?
<Kubuxu>
whyrusleeping: ^^
nicolagreco has joined #ipfs
<whyrusleeping>
Kubuxu: off the top of my head 'ipfs refs local | grep <hash>'
<whyrusleeping>
there was going to be some work towards adding a `--local` flag that would only check the local objects
<nicolagreco>
is ipfs mount implemented via fuse?
<Kubuxu>
Hmm, not optimal if I waned to do it from site/HTTP request.
<Kubuxu>
nicolagreco: yes
<whyrusleeping>
nicolagreco: yes
<whyrusleeping>
Kubuxu: yeah. It shouldnt be too difficult to PR a `--local` flag into a command to do what you want
<whyrusleeping>
would be pretty nice to havve
<nicolagreco>
perfect - so users havae to install fuse outside of ipfs?
zjohnson has quit [Ping timeout: 260 seconds]
<Kubuxu>
Yes they have to have FUSE userspace.
<ipfsbot>
[js-ipfs-api] diasdavid pushed 2 new commits to master: https://git.io/vzrBz
<ipfsbot>
js-ipfs-api/master c8aeadd David Dias: chore: build
<ipfsbot>
js-ipfs-api/master 934e869 David Dias: chore: release version v2.11.0
<Kubuxu>
What command would it be?
spikebike has quit [Ping timeout: 265 seconds]
<whyrusleeping>
Kubuxu: i'd like to have that flag on a lot of different commands
<grncdr>
ah, somehow I got the wrong impression from a different issue
<lgierth>
oh i think to be fair it *does* exist
<lgierth>
but in a weird wibbly-wobbly state
<lgierth>
you write by POSTing, and that pr ^ is testament to us changing our minds about it
<grncdr>
got it
<grncdr>
in my case, I'm really more waiting on https://github.com/ipfs/infrastructure/issues/105, but thought maybe I could play around with it locally first. atm it's not blocking me though so I'll wait a bit
parkan has quit [Ping timeout: 265 seconds]
jaboja has joined #ipfs
<lgierth>
grncdr: yep that one's not gonna happen before the PR merges
joshbuddy has joined #ipfs
chriscool has quit [Quit: Leaving.]
chriscool has joined #ipfs
chriscool has quit [Client Quit]
chriscool has joined #ipfs
jaboja has quit [Ping timeout: 264 seconds]
anticore has quit [Quit: bye]
pfraze has joined #ipfs
pfraze has quit [Remote host closed the connection]
Matoro has joined #ipfs
joshbuddy has quit [Quit: joshbuddy]
Encrypt has joined #ipfs
joshbuddy has joined #ipfs
pfraze has joined #ipfs
adamc1999 has quit [Ping timeout: 256 seconds]
rendar has quit [Ping timeout: 265 seconds]
Encrypt has quit [Read error: Connection reset by peer]
Encrypt has joined #ipfs
jaboja has joined #ipfs
rendar has joined #ipfs
pfraze has quit [Remote host closed the connection]
voxelot has quit [Ping timeout: 240 seconds]
neurosis12 has quit [Remote host closed the connection]
computerfreak has quit [Quit: Leaving.]
voxelot has joined #ipfs
jaboja has quit [Remote host closed the connection]
Kubuxu has quit [Remote host closed the connection]