#ipfs

/

      • maxlath1 has quit
      • vtomole joined the channel
      • ianopolous has quit
      • sametsisartenep has quit
      • herzmeisterderw joined the channel
      • hoboprimate has quit
      • hoboprimate joined the channel
      • Mateon1 has quit
      • herzmeister has quit
      • baikal has quit
      • baikal joined the channel
      • infinity0 has quit
      • infinity0 joined the channel
      • level56 joined the channel
      • kevina
        whyrusleeping: you around?
      • anonymuse joined the channel
      • captain_morgan joined the channel
      • anewuser joined the channel
      • Kane` joined the channel
      • LeerSpace has quit
      • fleeky__ joined the channel
      • whyrusleeping
        kevina: yeah, am now
      • kevina
        I'm afraid https://github.com/ipfs/go-ipfs/pull/3300 needs another look, see my comments
      • (basically I am not liking the "tar like" behavior model)
      • whyrusleeping: ^
      • fleeky_ has quit
      • f4_ joined the channel
      • f4_ has quit
      • f4_ joined the channel
      • f4_ joined the channel
      • whyrusleeping
        kevina: responded
      • f4_ joined the channel
      • f4_ joined the channel
      • f4_ joined the channel
      • f4_ has quit
      • captain_morgan has quit
      • ianopolous joined the channel
      • kevina
        whyrusleeping: responded
      • chris6131 joined the channel
      • chris613 has quit
      • whyrusleeping: we have a problem. Basically what you are trying to propose is to use "-w" by default now. This is the exact opposite of what I was trying to do with #3258.
      • whyrusleeping
        I think an argument could be made both ways, but jbenet and I discussed this and decided that having 'one add per pin' makes the most sense
      • its consistent, and simple
      • if you want to add and pin two directories separately, make two calls
      • kevina
        whyrusleeping: add calls are really expensive when adding lots of files
      • whyrusleeping
        no, not really
      • kevina
        whyrusleeping: I know because I tried
      • whyrusleeping
        'ipfs add X Y' and 'ipfs add X && ipfs add Y' should be roughly the same speed
      • the issue right now is that pinning is slow
      • kevina
        they are not, and it is not related to pinning at all.
      • whyrusleeping: basically it sounds like a decision has been made and this is a done deal
      • whyrusleeping
        i see no reason why those two should have different performance characteristics
      • if there is such a discrepancy, thats a bug
      • and we should address that as such
      • kevina
        whyrusleeping: trust me they do
      • whyrusleeping
        as opposed to using that to force our decisions and make other apis more difficult
      • kevina
        whyrusleeping: and what about add "-q"?
      • whyrusleeping
        yeah, that presents an interesting problem
      • kevina
        whyrusleeping: I think this needs more careful thought.
      • whyrusleeping
        the output doesnt necessarily have to always have the directory hashes
      • kevina
        whyrusleeping: but you seam to think this a done deal
      • whyrusleeping
        for the most part, yeah.
      • kevina
        whyrusleeping: I would like others input before we make a decision like that
      • whyrusleeping: but I am not in charge
      • whyrusleeping
        We want to mimic the behaviour of tar as much as we can in this
      • sure, make an issue and lets describe each of the usecases
      • and expected behaviours
      • we can weigh on the tradeoffs there
      • anewuser has quit
      • kevina
        whyrusleeping: thank you, I hope I was not coming across as aggressive
      • whyrusleeping
        nah, itsall good :)
      • you gotta fight for these sorts of things, its appreciated
      • kevina
        thanks.
      • I will create an issue and I will also post some performance stats on adding files individually
      • this entire thing came about because adding files one at a time was so slow.
      • Mitar has quit
      • as that is how I handle adding files for the filestore and keeping a directory in sync
      • lgierth
        rht had this great add benchmark script
      • which compared to git, tar, zip, etc.
      • Mitar joined the channel
      • would be really cool to get a hand on that script and add ipfs variants for the different stores
      • jeff-the-killer has quit
      • jeffl35 joined the channel
      • kevina
        lgierth: what we are comparing is the cost of adding a files one command at time to the cost of adding multiple files in the same command
      • lgierth: I don't think that is what you had in mind, apologies if I was mistaken.
      • whyrusleeping
        kevina: making 'adding files one at a time' faster is one of our big goals this quarter
      • kevina
        whyrusleeping: the difference is *huge* at least an order of magnitude.
      • herzmeisterderw has quit
      • herzmeister joined the channel
      • whyrusleeping
        yeap, we're going to investigate that
      • captain_morgan joined the channel
      • lgierth
        kevina: yeah that would be part of the benchmark suite too -- measure different ways of adding files, on different tools (variants of ipfs, and things like zip, tar, git)
      • then aim for getting really good in each graph
      • then profit
      • chax has quit
      • DiCE1904 has quit
      • DiCE1904 joined the channel
      • lapinot has quit
      • lapinot joined the channel
      • captain_morgan has quit
      • HostFat_ has quit
      • sdiver has quit
      • achin is about to OD on egg nog and extra halloween candy
      • kevina
        whyrusleeping: if we go your way with "tar like" behavior are you against a different command that _just_ adds (and pins, but only to prevent GC) files?
      • whyrusleeping: and this new command can add multiple files at a time, but never directories
      • that is what I need for the filestore for now
      • whyrusleeping
        kevina: hrm... why?
      • i'm not sure what the usecase there is
      • chungy has quit
      • kevina
        that is how a directory of files will be kept in sync with the filestore
      • whyrusleeping
        i'm not sure i follow
      • kevina
        as files change they are readded individually
      • what I really want is to separating the adding of files with the building of directories.
      • teenytree has quit
      • sdiver joined the channel
      • whyrusleeping: basically you add a large directory, then a few files change
      • captain_morgan joined the channel
      • whyrusleeping
        okay
      • kevina
        ... it would be a complete waste of time to readd the entire directory
      • whyrusleeping
        so, your main concern with 'one pin per add' is that its too expensive to add files individually?
      • kevina
        yes
      • and I don't want the creation of directory objects when adding a file using an absolute path
      • whyrusleeping
        okay, so if i told you that i will magically make that better, would you still be against it?
      • did you write up an issue? i'll add a few comments there if so
      • kevina
        Not yet, give me a few minutes....
      • whyrusleeping
        okay
      • arpu joined the channel
      • Once you do i'll post my compromise on it
      • ygrek joined the channel
      • kevina
      • I didn't spell out the issues yet but I will let you go first.
      • chungy joined the channel
      • anonymuse joined the channel
      • whyrusleeping: ^
      • captain_morgan has quit
      • whyrusleeping
        kevina: on it
      • lgierth
        whyrusleeping: i like your cli proposal for add
      • or example or whatever
      • i like it
      • whyrusleeping
        lgierth: thanks :)
      • its my compromise
      • kevina
        whyrusleeping: our messages crossed :)
      • I think.
      • whyrusleeping
        kevina: cool cool, think that should work out?
      • any other questions i can answer while i'm in the mindset?
      • lgierth
        typing `reboot` in the wrong terminal: priceless
      • whyrusleeping
        lgierth: 👌
      • kevina
        lgierth: a very good reason for sudo with a password :)
      • lgierth
        nevermind, just rebooted my computer instead of that vps
      • mh no, somehow i didn't need sudo