#ipfs

/

      • mguentner has quit
      • mguentner joined the channel
      • SuperPhly has quit
      • se3000 has quit
      • tmg has quit
      • anbora has quit
      • se3000 joined the channel
      • shizy joined the channel
      • whyrusleeping
        daviddias: current go-cid/go-multibase only supports base58
      • once we update it (0.4.6) it will support other encodings
      • SchrodingersScat
        kpcyrd: what's the service?
      • chris613 joined the channel
      • wmoh has quit
      • shizy has quit
      • aquentson1 joined the channel
      • se3000 joined the channel
      • aquentson has quit
      • warner joined the channel
      • matoro joined the channel
      • fleeky_ has quit
      • fleeky__ joined the channel
      • aeternity has quit
      • wfjsw[m] joined the channel
      • maciej_ has quit
      • cemerick joined the channel
      • kpcyrd
        SchrodingersScat: https://ipfs.ink
      • I wrote a frontend build system, I'm testing if my docker container still builds and merge it to master in some minutes
      • cemerick joined the channel
      • SchrodingersScat
        I see
      • Hoosilon joined the channel
      • infinity0_ joined the channel
      • infinity0_ has quit
      • infinity0_ joined the channel
      • infinity0_ is now known as infinity0
      • chris6131 joined the channel
      • galois_dmz has quit
      • wallacoloo____ has quit
      • infinity0 joined the channel
      • arpu has quit
      • chris613 has quit
      • infinity0 joined the channel
      • infinity0 joined the channel
      • dryajov1 joined the channel
      • infinity0 joined the channel
      • infinity0 joined the channel
      • infinity0 joined the channel
      • arpu joined the channel
      • ckwaldon has quit
      • lapinot has quit
      • ckwaldon joined the channel
      • lapinot joined the channel
      • rplevy[m] joined the channel
      • ploop joined the channel
      • sk23[m] joined the channel
      • HostFat_ joined the channel
      • dryajov1 has quit
      • pjz
        gah, is v0.4.5 out yet?
      • oh, buh, somehow I'm running -pre2
      • hmm
      • dryajov1 joined the channel
      • I suspect my ipfs-update script isn't discriminatory :)
      • wmoh joined the channel
      • aquentson1 has quit
      • aquentson1 joined the channel
      • se3000 has quit
      • wallacoloo____ joined the channel
      • stevenaleach joined the channel
      • wallacoloo____ has left the channel
      • wallacoloo____ joined the channel
      • Boomerang has quit
      • Foxcool__ joined the channel
      • _whitelogger joined the channel
      • SuperPhly joined the channel
      • SuperPhly has quit
      • vtomole joined the channel
      • mguentner has quit
      • mguentner joined the channel
      • what's the default hash that ipfs uses?
      • ah, sha256
      • though: given a pinned object, is it accessible only via the hash it was pinned with? or are the other hashes calculated and usable as lookup also?
      • muvlon_
        what other hashes do you mean?
      • appa has quit
      • aaaaaaaaa____
        kpcyrd: that looks neat
      • mguentner2 joined the channel
      • pjz
        muvlon_: IPFS uses a multihash for its hash addresses. A multihash is a wrapper for (you guessed it) multiple hash functions
      • mguentner has quit
      • muvlon_: so a single block of data has multiple multihashes, one per hash function (sha1, sha256, etc)
      • kpcyrd: cute! I'm trying to figure out how to do something similar, but allow uploading files (size limited!)
      • kpcyrd: are you pinning the resultant docs? if so, for how long?
      • muvlon_
        pjz: are these actually computet right now?
      • pjz
        muvlon_: I don't know; that's what I'm asking :)
      • vtomole
        Hitting a bit of a creative blcok on the design of an IPOS. What is the difference between "boot a VM from an image in ipfs" and "boot a VM from a filesystem tree in ipfs"? Illustrated here:https://github.com/ipfs/go-ipfs/blob/maste...
      • pjz
        muvlon_: clearly the sha256 one is, as it's used for the address - I'm wondering if the others are computed but not shown or just not computed
      • vtomole: deduplication
      • vtomole: change an image and the whole file changes. (well, the whole file after the point of the change)
      • vtomole: change a file in an fs tree in ipfs and only that branch of the fs changes (well, and the parent hashes up to the root)
      • muvlon_
        pjz: ipfs already has rabin chunking though, so hopefully the whole image wouldn't need to be re-added to ipfs, right?
      • vtomole
        muvlon_: sadly IPOS is still in that primitive stage where you ahve to readd the fileafter you've changed it
      • *re-add
      • pjz
        muvlon_: depends on if your images are fixed-size
      • muvlon_
        they shouldn't need to be fixed size for rabin chunking to work
      • vtomole
        But the chunking is efficient. Not efficient enough though :(, adding a whole os (1+ GB) takes a while.
      • muvlon_
        but I haven't looked closely at ipfs' implementation of it
      • pjz
        muvlon_: ipfs does rabin chunking? I thought they were fixed-size chunks
      • muvlon_: on disk they're all 256k chunks by default
      • muvlon_: ... so not rabin chunked
      • arkimedes joined the channel
      • oh, hm, I guess they *could* still be, given the probably lack of overlap in my dataset
      • tmg joined the channel
      • vtomole
        Seems like booting from a filesystem tree would be kinda compilated. If i ipfs add the file system i changed, how would that change be prevelant in the vm image?
      • muvlon_
        I'd boot an initial ramdisk from an image, then load a rootfs as a tree
      • pjz
        sure, initrd's are usually image files anyway
      • they're small and don't change often
      • bootstrappy things
      • in fact, you could do the whole ipos thing as an initrd that was pxe-bootable across the internet
      • muvlon_ would love to see an EFI implementation of ipfs
      • ...from an ipfs gateway, of course :)
      • chris6131 has quit
      • muvlon_: I think a minimal bootdisk/PXE-bootable image would be pretty close
      • vtomole
        Would that solve the problem though? (avoiding having to"ipfs get" 1GB+) everytime i want to use my OS on a different computer?
      • mildred_ joined the channel
      • aquentson joined the channel
      • mildred has quit
      • aquentson1 has quit
      • I guess it's not a big deal now.
      • wallacoloo____ has quit
      • pjz
        I dunno; I'm not entirely sold on the IPOS idea in the first place :)
      • vtomole
        What do you think is wrong with it?
      • wallacoloo____ joined the channel
      • pjz
        I think ipfs is best suited for 'broadcasting' files: one sender, multiple receivers. IPOS... isn't that.
      • I might be wrong: you might be able to do like docker and figure out how to layer your deltas on top of an underlying OS image that lots of people have cached, and therefore contribute bandwidth to.
      • even then, though, it's not going to boot *quickly*
      • because, like you said, you have to d/l a gig or so of OS before you can boot
      • ah! found the public PXE site I was looking for: http://boot.salstar.sk/
      • vtomole
        I haven't reached out to jbenet, but i think this is how he wants to design his IPOS. I don't know though. If he's online i would love to hear his thoughts :)
      • muvlon_
        pjz: isn't "one file, multiple receiver" just what a distro needs?
      • maybe not for the initrd but for package management
      • pjz
        muvlon_: ...sure, and I'm all for it for package management. But do you want to have to wait for the entire OS-worth of packages to install every time you boot?
      • vtomole
        muvlon_ we want vms to be personal. So different files for different people
      • muvlon_
        yes sure
      • but, for example, the entire /usr directory can be ipfs
      • pjz
        has anyone worked on apt-ipfs?
      • muvlon_
        there's nothing personal in there
      • vtomole
        caching is our only hope then..
      • pjz
        er, apt-transport-ipfs
      • vtomole
        pjz you mean like gx?
      • f[x] has quit
      • muvlon_
        or, if you abandon FHS in favor of something like objectroot, everything under /org could be ipfs
      • pjz
        sure, that's a point: integrate ipfs into NixOS
      • it has the entire fs as links to hashes anyway