Video Of The Week: IPFS and Filecoin

Our portfolio company Protocol Labs built IPFS and will now launch a token for decentralized storage called Filecoin.

Juan Benet, the founder of Protocol Labs, explains all of that in this Y Combinator video.

This is super long, but the first part (30-40 mins) explains most of what they are up to.


Comments (Archived):

  1. thepcmate

    The first 30-40 minutes are just awesome. Thanks dude

  2. jason wright

    just awesome…for the US domiciled accredited investor class.

  3. Frank W. Miller

    Couple of technicalish comments/questions. They’re looking at interesting problems but I couldn’t help wonder how much history he has studied on this stuff.He said its time to move to content vs location addressing. That may be true for whatever his app is but the Internet can support both so it not an either or.Human readability? I couldn’t figure out what he was actually trying to name. It sounded at the end like it was making the keys for the content addressing human readable. I’m not sure why this is important? If the keys are the content then whether you can do that depends on whether the content is human readable. For example, if the content is a binary encoded video, it will be tuff to come up with anything human readable to use for the key.Proof of storage? This is a little tongue in cheek but proofs are something grad students do when they are coming up with new algorithms. Executing proofs online seems to me to be of dubious value and in the case of bitcoin, downright evil. So now rather than a building full of servers being built next to dams to supply the power, we’ll have multi acre installations filled with SSDs and hard drives? The proof-of-xxx thing is a problem, not a solution.And then there is a sort of general comment. There seems to be a lot of interest from you (Fred) and many others in fully decentralized apps (it has a Silicon Valleyish feel to it actually). I see things like Chordish based p2p networks tracing back to variations on consistent hashing as the fundamental technology in a recurring way on your site. These are great things and Skype is great example of an app that was (at least originally) fully distributed. But this thing I’ve been referring to, the potential for serious catastrophic chains of events occurring (the Ethereum thing a month or so ago is an example) is inherent in their designs. Full distribution yields no governance which makes thing unworkable because of the USERS behaviors, not the computers. Think of it this way. If you’re a hacker, you really really want networks to be fully distributed. If you’re able to break in, there’s noone to stop you. That fundamental human behavior is imo why fully distributed protocols have had limited practical and commercial success of the last 50-60 years (which is how long folks have been banging on these same problems).

    1. Greg Kieser

      Hmmm… executing proofs with bitcoin seem downright evil? Tell us more.Also, when you say “no governance” you mean no centrally chosen governance, right? With blockchain based apps, the participants in the network ultimately choose the governance model and leaders, by choosing whether or not to run the app or by choosing between one fork and another. No?

      1. Frank W. Miller

        As I’ve talked about before. Proof of work for bitcoin is a contrivance. It not only serves no purpose but to simulate scarcity but it will suck gigajoules of energy out of the grid that could be used for, oh, growing crops, or powering homes and cars and businesses or something else that actually is useful.I’m referring to operational governance. If I understand it correctly, when a transaction completes, there’s no way to undo it with bitcoin (primarily because it is a fully distributed system), even tho it may have been malicious. Ethereum at least had the technical ability to roll things back to some degree (from what I read anyway). How they ultimately use that technical capability is something they are still wrestling with I would imagine. My opinion is they should embrace it and establish policies and rules for how it will be used. Its a good thing and it will give users more confidence in using Ethereum.

  4. Vendita Auto

    Thanks, not a techie but good to open my horizon. So often reminded of the Rovelli quote:”Time is information we don’t have.””Time is our ignorance.”

  5. Richard

    I often wonder how technically challenging things like IPFS actullay are? Is the novelty in the idea or the algorithms that make up the code?IPFS sounding like a cool research project, but how many phd students in universities are working on similar projects? what is the value proposition for distributed storage? Who are the first customers who would want to pay for this? What are the risk reward issues? I would think that any large company would prohibit access to any data stored on random hard drives spread throughout the world?

    1. sigmaalgebra

      > IPFS sounding like a cool research project, but how many phd students in universities are working on similar projects?Let’s see: (A) Graduate students are silly, time-wasting, impractical, fools engaging in intellectual self-abuse. (B) Graduate students have nothing better to do than waste time on projects like IPFS. (C) Graduate students are so plentiful that likely many of them around the world are working on projects like IPFS. (D) Since graduate students have no sense of what is practical, they are plenty eager to work on solutions still looking for problems. (E) Also graduate students are so isolated and disconnected that they would not be able to find practical problems to work on. (F) Graduate students are highly inferior, laughably so, to us hard nosed, down and dirty, reality based, people in the world of business.Do I have that about right?Actually, graduate students are expected to work on research which is supposed to be “new, correct, and significant”. Working on something like IPFS, which already exists, would not be new and even if new might not be significant.Further, at better research universities, in computer science, the goal is to find the fundamental of the field. Just writing some code for some practical functionality, even if new, does not always qualify as “significant” or “fundamental”.Running down graduate students, graduate schools, academics, and related research is not wise. Instead, a huge fraction of the best progress in our civilization and economy is dependent on good academic research, dependent in the sense that, likely, if the research had not been done then the progress would not have happened.E.g., academic work on, say, the photoelectric effect, e.g., by Einstein, led to quantum mechanics and much better understanding of how solids and conductors and semi-conductors work. Before WWII Bell Labs saw this and saw the potential for a solid state amplifier to replace hot, expensive, noisy, unreliable vacuum tubes. Soon after WWII, Drs. Shockley, Bardeen, and Brattain had a working transistor. Bell Labs saw the value right away and arranged to give the patent rights to the world for free. If Bell Labs had had a royalty of only a penny per transistor, then by now they might have owned the world and everything in it — just do some simple arithmetic. That’s one example. Our economy is awash in more examples from pure/applied math, physical science, biology, and medical science. E.g., yesterday I was mowing the backyard and saw some ugly weeds, dark green with leaves that looked like serrated knife edges, some weeds with leaves maybe 1 square foot in area, and some poison ivy. So, I got out my spray bottle with some Roundup solution and give a few squirts.Just now, with cheap computing and the Internet, information technology is wide open to astounding, and astoundingly valuable, applications of original pure/applied math, that’s math and not computer science, with the foundations going back 100+ years in academics. .

  6. creative group

    CONTRIBUTORS:OFF TOPIC ALERT! BUT ON TOPICIRS Says It Will Limit Bitcoin Audits…Does a person really want to provide information that can be available to the IRS for no other reason than to fish for what they view as tax avoidance but really an opportunity at profit sharing. You had to know when smart money who can write a check to offset any profits (But hates it with a passion, the anonymous requests to stop inquiry would read a whos whos list) tt he IRS would be at the door.

  7. sigmaalgebra

    IPFS? Sure, as a founder of a Web site startup, I want to keep up on such things. Maybe something about IPFS could help my startup or Web site?There is a description of IPFS at…By the way, apparently IPFS is supposed to abbreviate InterPlanetary File System.To me, IPFS looks like a solution still looking for a significant problem it is good at solving.IPFS is supposed to help the Web — (A) for servers, lower bandwidth costs, help stop distributed denial of service attacks (DDOS), etc. and (B) for the users make the Web faster, especially reduce latency, and more reliable.As I understand it, IPFS takes an ordinary file, breaks it into pieces, based on a hash (what happens with hash collisions I didn’t see), gives each piece a name unique for all time, and then stores the pieces possibly on many different computer systems in many geographically distributed places.So, there is a big assumption here: The files stored by IPFS don’t change over time. That is, in terminology now common in parts of computing, the files are immutable. So, IPFS is for files that don’t change.Sure, if a file does change, say, from a revision of some kind, then maybe go ahead and call the revised version a new file and maybe reuse the parts of the old file that did not change.Hmm …. Files that don’t change: Will that do much to help the Web?From what I’ve read, for one use for the Web, IPFS is supposed to help at the level HTTP — hypertext transfer protocol.I’m no grand expert on IPFS, HTTP, or HTML (hyper-text markup language). For all the Web pages for my startup all the code appears to work, but I didn’t dig much deeper into HTTP and HTML than I needed to get the pages to work, and I didn’t write even a single line of JavaScript.But for IPFS and the Web, it looks like there are some flaws in some of the claims. I review the situation as I understand it:(1) HTTP.As I understand HTTP, a user enters the IP address of a Web server or enters a domain name which by the domain name system (DNS) is translated to an IP address. Then the user’s Web browser uses standard TCP/IP to create a TCP/IP socket connection with the Web server with the IP address and sends the Web server an HTTP request GET to get the default or home Web page for that Web site.The Web server, then, on the same TCP/IP socket, returns just some one file. As for standard e-mail, the file starts with some header lines, has a blank line, and then has the message body which for the Web is some or all of the Web page the user was requesting.Maybe as in e-mail and (the clear as mud, awash in undefined terminology, horribly badly written)…there can be some images, sound, included (likely using much or all of the main means of multi-media internet mail extensions (MIME)).Okay. With the user’s HTTP GET request to the Web server, only that one file is sent to the user by the server.Commonly that file is unique in all the world for all time, e.g., may have time stamps, hidden fields with unique data, etc., and stands never to be sent again. Moreover, that file (1) had to have been created by the Web server AFTER the user’s HTTP GET request to the Web server and (2) stands never to be used or sent again. Then for that file IPFS would be useless.Next, sure, commonly when a user gets a Web page from a Web server, the user also gets dozens of files of CSS (cascading style sheet), JavaScript, images, etc. Some of the images are for the main content of the Web page, and more images can be for ads.As I understand the protocols, of those dozens of files, commonly the file names and their URLs are not clear just from the HTTP format, syntax, etc. but are from processing the file sent via HTTP.(2) HTML.That processing is, first, just via HTML. So, the file sent via HTTP is parsed by the user’s Web browser according to the syntax of HTML. From that parsing, the URLs for the files for CSS, JavaScript, images, etc. can be read.(3) JavaScriptNext, when files of JavaScript have been downloaded by the user’s Web browser, the JavaScript code executed by the user’s Web browser can request more files.Next, once the resulting Web page is presented to the user, user actions can cause JavaScript to download still more files.(4) Unique Pages.Next, as the user makes use of the Web site, they can see many Web pages, but commonly each of those pages is unique to that user, time, and the user’s logical session with the Web site and its user interface and user experience so that the pages have to come from the Web server and not just from IPFS.(5) User Interactions.Next, usually the user will be interacting with the Web site, and the Web server side of those interactions will have to be from what is logically the one Web server itself, and that is difficult to distribute. So, IPFS does not help the Web site have a distributed Web server.But, sure, for those dozens of files for HTML, CSS, JavaScript, images, etc., maybe they change slowly and could come from IPFS instead of directly from the Web server. So, that’s a case of caching files that don’t change.(6) CDNs and AkamaiOkay, but such file caching is an old idea for the Web and the main reason for content distribution networks (CDNs) and IIRC much of the business of Akamai.In particular, IIRC for the issues of latency and data rate on the Internet backbone, say, someone in the third world using a Web site in Manhattan with several Internet backbone links involved, Akamai has some distributed servers to reduce that latency and data rate on the backbone.I’m reluctant to see that IPFS will play a role in improving such file caching.(7) Do Web Sites Care?Next, on the need, from some of what I’ve seen on the Internet, Web site operators are not very concerned about Internet data rate or latency.Why? Two reasons:(A) It’s easy enough to include any and all JavaScript and CSS code in the HTML file, just using HTML syntax and not HTTP syntax, first sent to the user by the Web server in response to the user’s first HTTP GET request to the Web server.E.g., at my Web site, no separate CSS files are sent. Microsoft’s ASP (active server pages) .NET that I am using has plenty of functionality for having the CSS files as separate files on the file system of my Web server, but I deliberately do not make use of that functionality. Instead, I just include the CSS data in the first HTML file.For JavaScript, I didn’t write any. Microsoft’s ASP.NET wrote a little JavaScript for me — it’s short and looks nice, but I haven’t bothered to read it. I believe that that JavaScript is to support some features of ASP.NET that could not be obtained with just HTML. Okay. So, I’ve seen that Microsoft’s software does the right things with that JavaScript; I just ignore it.(B) It’s common for Web sites, e.g., Business Insider, just to send with every Web page nearly everything the Web site has and is using that day or week or some such.E.g., when download a page from Business Insider and look at the files that get, see files for what appears to be all the CSS, JavaScript, and images they are using for their Web site.It would be easy enough for that site to send with each page only the files needed by that page, but apparently the Internet data rate charges are so low that they do not bother to save the money.So, the savings indicated for IPFS some Web sites just don’t much care about.(8) Conclusions.It appears that, as the IPFS people are looking for applications of their technology and are trying to say that IPFS can save data rate and latency on the Web, CDNs, Akamai, how Web sites write their Web pages, how much (only a little) those sites care about the savings, and the fact that such a large fraction of Web pages are created only just before they are sent and never sent again it appears that IPFS has not much to offer Web sites.

  8. Nate George

    Fred, what’re your thoughts on Metalpay?

  9. William Mougayar

    IPFS is a perfect example of a vertical/special purpose protocol. What is not so well known is that it is not exclusive to blockchains.

  10. pinoytamb

    If you are keen interested and biggest fan of pinoy tv you must to watch and visit the very famous and quick start website that is really best for your mobile phone because of its Mobile friendly setup with less advertisement and ISO Certified

  11. pinoytamb

    Proof of storage? This is a pinoy tambayan tongue in cheek but proofs are something grad students do when they are coming up with new algorithms. Executing proofs online seems to me to be of dubious value and in the case of bitcoin, downright evil. So now rather than a building full of servers being built next to dams to supply the powe