#ipfs

/

      • dawuud
        makes sense
      • voxelot has quit
      • Eudaimonstro has quit
      • Eudaimonstro joined the channel
      • mvollrath joined the channel
      • xelra joined the channel
      • jbenet
        Wanderers: please watch https://vimeo.com/108650530
      • patcon joined the channel
      • !pin /ipfs/QmWbEAcGz2KWEpkRcLaVX2KMitohmNMxiKxshUsYcsc62G
      • pinbot
        now pinning /ipfs/QmWbEAcGz2KWEpkRcLaVX2KMitohmNMxiKxshUsYcsc62G
      • notduncansmith joined the channel
      • notduncansmith has quit
      • [host 3] failed to grab refs for /ipfs/QmWbEAcGz2KWEpkRcLaVX2KMitohmNMxiKxshUsYcsc62G: Post http://[fc4e:5427:3cd0:cc4c:4770:25bb:a682:d06c...: dial tcp [fc4e:5427:3cd0:cc4c:4770:25bb:a682:d06c]:5001: connection timed out
      • jbenet
        whyrusleeping why do _offline_ operations take a million years now?
      • (like w/o a daemon)
      • is it ttrying to bootstrap for whatev reason?
      • wking around? mind providing /ipfs/QmeyrsAMNmU5oXEkLth8YqAiPsZubhg4ou8WbaR5Pj2E8F ? looks like it didnt get pinned or something
      • lgierth
        hrm that one node has been failing a couple of time over the past days i think
      • its cjdns routing table is in a weird state
      • (restarting)
      • notduncansmith joined the channel
      • notduncansmith has quit
      • voxelot joined the channel
      • notduncansmith joined the channel
      • notduncansmith has quit
      • jbenet
        whyrusleeping i cant CR mfs. cant even run it
      • therealplato has quit
      • fiatjaf joined the channel
      • fiatjaf has left the channel
      • whyrusleeping
        jbenet: huh, that PR for krl broke it... whoops
      • jbenet
        whyrusleeping: may want to fix those things (like rebasing) this round
      • whyrusleeping
        yep, working on it now
      • notduncansmith joined the channel
      • notduncansmith has quit
      • wking
        jbenet: I just spun up a daemon serving serving /ipfs/QmeyrsAMNmU5oXEkLth8YqAiPsZubhg4ou8WbaR5Pj2E8F , but see also https://groups.google.com/a/opencontainers.org/... for a quick demo and notes about the backing Git repository for folks who want to add that content themselves
      • notduncansmith joined the channel
      • notduncansmith has quit
      • jbenet
        wking thanks!
      • !pin QmeyrsAMNmU5oXEkLth8YqAiPsZubhg4ou8WbaR5Pj2E8F
      • pinbot
        now pinning /ipfs/QmeyrsAMNmU5oXEkLth8YqAiPsZubhg4ou8WbaR5Pj2E8F
      • jbenet
        !befriend wking
      • ah not sure that will work, may not be re-deployed yet
      • pinbot
        [host 3] failed to grab refs for /ipfs/QmeyrsAMNmU5oXEkLth8YqAiPsZubhg4ou8WbaR5Pj2E8F: Post http://[fc4e:5427:3cd0:cc4c:4770:25bb:a682:d06c...: dial tcp [fc4e:5427:3cd0:cc4c:4770:25bb:a682:d06c]:5001: connection timed out
      • wking
        hmm, what's that failure about^? I'm pretty sure I have the whole tree locally
      • jbenet
        wking oh wow lots of ipfs discussion
      • wking thanks for giving me the link, and describing the ipfs way.
      • wking
        Working hard to keep bundles content addressable ;)
      • jbenet
        wking: omg yes. thanks for fighting the good fight.
      • looks like making ipget is a priority here
      • the daemon is definitely awkward
      • whyrusleeping: wanted to merge mfs tonight. but i may go to sleep soon. how far are you?
      • ipfsbot
        [13go-ipfs] 15whyrusleeping pushed 1 new commit to 06feat/mfs: 02http://git.io/vGg3M
      • 13go-ipfs/06feat/mfs 144eaf136 15Jeromy: fix issue caused by lazy creation of mfs...
      • whyrusleeping
        jbenet: it works now
      • gotta get the other CR stuff
      • jbenet
        wking: the failure is one node failing
      • notduncansmith joined the channel
      • Eudaimonstro has quit
      • whyrusleeping
        jbenet: adding examples to each commands doc, then pushing
      • notduncansmith joined the channel
      • jbenet: theres some comments!
      • zignig
        o?
      • o/ all
      • whyrusleeping
        zignig: hey!
      • zignig is back on a tectonic plate.
      • zignig
        whyrusleeping: waht's cooking ?
      • whyrusleeping
        jbenet: whoops... messed that up
      • theeeere we go
      • ipfsbot
        [13go-ipfs] 15whyrusleeping pushed 2 new commits to 06feat/mfs: 02http://git.io/vGgcQ
      • 13go-ipfs/06feat/mfs 14d75ad43 15Jeromy: update comments...
      • 13go-ipfs/06feat/mfs 1428d0392 15Jeromy: add examples to docs...
      • whyrusleeping
        aaand theres more CR i missed
      • okket joined the channel
      • jbenet
        whyrusleeping: is write fixed? cant use it
      • there's your UX feedback
      • zignig you werent on one?
      • zignig i thought you were on an oil rig-- was it floating free?
      • whyrusleeping: " i do want to keep the top level mounts as 'special' objects." why?
      • whyrusleeping
        jbenet: where the hell is '-c' coming from?
      • jbenet
        whyrusleeping root, i think it's for alternate config
      • not sure.
      • whyrusleeping
        jbenet: because it makes it easier to do things like mount an ipns entry inside mfs
      • mfs is a set of mountpoints
      • jbenet
        mountpoints in unix can be mounted anywhere
      • not just root
      • the mountpoints could be ipfs objects available anywhere
      • whyrusleeping
        later, the create command could do things like 'ipfs mfs create --ipns=keyname ipnsfs'
      • jbenet
        ahhh this is what dm was talking about
      • "doing symlinks right"
      • "symlinks" being ipns names for us
      • he said this was hard at the time.
      • notduncansmith joined the channel
      • yeah, so i think if we have an ipfs object that links to an ipns name, it can be like the "mount point" you expect
      • whyrusleeping
        and maybe i want 'ipfs mfs create --pubcmd=/bin/publish-object --close-cmd=/bin/do-this-on-close myweirdmount'
      • jbenet
        what does that mean?
      • whyrusleeping
        every time a change propogates to the root, it could run '/bin/publish-object <root hash>'
      • jbenet
        that's pretty cool
      • like git hooks
      • whyrusleeping
        and when its closed/unmounted, would run '/bin/do-this-on-close <root hash>'
      • yep
      • jbenet
        wonder if that could be data on the object itself
      • should*
      • like the mount point object
      • suppose i could "ipfs mfs mount --mutate-hook=/ipfs/<hooks>/ipns-publish-hook.sh /ipns/<key> /files/foo/bar/baz/name"
      • then i could "ipfs mfs write /files/foo/bar/baz/name/a <a && ipfs mfs write /files/foo/bar/baz/name/b <b && ipfs mfs write /files/foo/c <c"
      • and the hook would only run for the first 2 commands
      • whyrusleeping
        the publish calls are currently coalesced to prevent super slowness
      • jbenet
        (sure that's fine)
      • (that's for your "weirdmount" example above, but with unixy mountability)
      • whyrusleeping
        yeah
      • i still like to think of each 'mfs root' as a separate filesystem.
      • i like being able to have multiple different filesystems
      • jbenet
        i dont understand what that gives you
      • it limits you
      • whyrusleeping
        separation?
      • jbenet
        things are separated equally
      • it may be virtual separation "it feels different"
      • whyrusleeping
        not doing it limits you to a single root, youre all under a single namespace basically
      • jbenet
        but the underlying data structure is the same.
      • it already is under a single namespace. whether you do it or not, "sessions" are still under one namespace.
      • imagine linux did this, for only the root of `/`
      • like /* were "different filesystems"
      • whyrusleeping
        mmm....
      • jbenet
        and had you use a different set of syscalls to touch things in /* vs in /*/** -- because, you know, "/proc" is a different fs than "/home"
      • (that o/ doesnt even make sense, partitioning the syscalls for the roots from their children, instead of per "different type of fs"
      • whyrusleeping o/
      • (ignore the cp -rf, i meant cp -r)
      • whyrusleeping
        this one: 'mfs read /files/foo/bar/baz | sed -i s/l/r | mfs write /files/foo/bar/baz' you cant really do in linux
      • cat X | blah blah > X
      • lilsheep joined the channel
      • notduncansmith joined the channel
      • what do you expect to happen with 'mfs rm -rf /' ?
      • jbenet
        linux can with sponge.
      • (and minus the -i in sed)
      • whyrusleeping
        sponge?
      • jbenet
        with "mfs rm -rf /" i expect all roots to be removed, i.e. really just setting / to the null dir
      • sponge buffers stdin, then outputs to stdout
      • a way to get those same file edits to work-- the reason it doesnt work is efficient unix piping
      • whyrusleeping
        oh, thats totally a hack...
      • jbenet
        why, it makes sense.
      • the limitation is not there as a feature, it's there by consequence of piping efficiency. the first line may flush through to be written before the second one is read.
      • whyrusleeping
        what if: "cat X | sed s/a/biggerthana/g > X"
      • the write pointer will go faster than the read pointer
      • unless youre saying "i know this doesnt work, but it would be cool if it did"
      • at which point i agree