Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Decentralized Web Primer

gitbook.com

Post reply on HN
Print view View original

51–60 of 97 posts

Re: Decentralized Web Primer

#51

Post by hedora »

wanderingbort wrote: post #40
colemannugent wrote: post #14

Earlier quoted context omitted.

From what I gathered by reading the whitepaper mass censorship is extremely hard to accomplish. > Could ISPs throttle or block IPFS content, for instance? They wouldn't know what was IPFS and what was regular encrypted TCP, the only way I can tell IPFS traffic apart in wireshark is the port number. Once ISPs start blockings random encrypted connections people will get mad when their games or other applications stop w…

> They wouldn't know what was IPFS and what was regular encrypted TCP The extreme conclusion of state-sponsored censorship is not a black list or DPI. It is a white list. It is easy to restrict traffic to a small set of addresses that are rigorously monitored for any ability to proxy non-conforming traffic. Everything else would get dropped on the floor. IPFS and the like sit too high in the OSI layers to combat stat…

I’d think it would support grafting in chunks of the merkle dag via usb drives, etc. Otherwise, I don’t understand the claims about disconnected operation and breaking the dependency on the internet backbone.

Re: Decentralized Web Primer

#52

Post by saurik »

0xcde4c3db wrote: post #49
saurik wrote: post #28

Earlier quoted context omitted.

> It relies on the DNS and HTTP protocols. DNS is fundamentally centralized.

How so? I see how "the" DNS is centralized in ICANN's list of root servers, but from a technical standpoint isn't the use of that list just a convention?

By this definition of "centralized", nothing is centralized. For DNS to have any meaning--particularly in the way it is used by HTML--requires there to be a person whose job is to run a server that maps "name a human enjoys using" to "name that a computer can connect to" in a way that is stable over time. To the extent to which DNS "the protocol" (as opposed to DNS "the concept") is incapable of enforcing that everyone uses the same servers (as no protocol can force that; again: with this definition, nothing is centralized), we have been building out cryptographic mitigations into other protocols (such as TLS and X.509) that act as a form of checksum against DNS failing to be used to access the one shared reality. Yes: a user can opt to use a totally different "shadow Internet" by using different DNS root servers and different CA root servers, but that is true of any protocol.

Re: Decentralized Web Primer

#53

Post by nyolfen »

0xcde4c3db wrote: post #49
saurik wrote: post #28

Earlier quoted context omitted.

> It relies on the DNS and HTTP protocols. DNS is fundamentally centralized.

How so? I see how "the" DNS is centralized in ICANN's list of root servers, but from a technical standpoint isn't the use of that list just a convention?

"the" web uses "the" dns

Re: Decentralized Web Primer

#54

Post by bogomipz »

hedora wrote: post #50
bogomipz wrote: post #44

Earlier quoted context omitted.

>""The Decentralized Web" is the WWW system Tim Berners-Lee invented and that we're all using right now." I believe the context of "The Decentralized Web" here is meant to contrast the increasing centralization of content from FB, Google, Medium, login walls pay-walls etc.

Those sites are all part of the decentralized web. Port 80 on my laser printer, router and flower pot are too. Using old well-defined terms to describe new concepts just leads to confusion. It would be like writing a story called “the history of the industrial revolution” that’s really about 3d printers.

So out of the one sentence that I wrote you decided to ignore the part where I wrote that decentralization was referring to centralization of content? Another word for centralization is consolidation.

You might want to look up the term "walled garden":

https://www.technologyreview.com/s/512316/facebook-and-googl...

Re: Decentralized Web Primer

#55

Post by fiatjaf »

JetSpiegel wrote: post #36
fiatjaf wrote: post #32

Earlier quoted context omitted.

"use a P2P protocol instead". Actually, just like "the cloud" is just someone else's computer, "P2P" is just someone else's server.

P2P is not like the cloud. It's someone else's computer, but you don't know nor care what the exact machine you are connecting to because they are mostly untrusted.

I just made an analogy, didn't say it was the same thing as the cloud.

My point was that in a P2P architecture you must be a server, everyone else must be a server. And being a server is costly. There must be incentives for someone to be a server, and that's not easy. I actually don't remember a single P2P protocol or framework that has solved that problem yet. Perhaps Bitcoin -- for miners only, not for normal nodes.

Re: Decentralized Web Primer

#56

Post by fiatjaf »

loup-vaillant wrote: post #39
fiatjaf wrote: post #32

Earlier quoted context omitted.

"use a P2P protocol instead". Actually, just like "the cloud" is just someone else's computer, "P2P" is just someone else's server.

P2P is everyone else's server. Each peer pulls its own weight. This is different from relying on a single entity.

Relying on others is fine, the problem is when someone relies on you. Please see my answer to the other comment.

Re: Decentralized Web Primer

#57

Post by 0xcde4c3db »

saurik wrote: post #52
0xcde4c3db wrote: post #49

Earlier quoted context omitted.

How so? I see how "the" DNS is centralized in ICANN's list of root servers, but from a technical standpoint isn't the use of that list just a convention?

By this definition of "centralized", nothing is centralized. For DNS to have any meaning--particularly in the way it is used by HTML--requires there to be a person whose job is to run a server that maps "name a human enjoys using" to "name that a computer can connect to" in a way that is stable over time. To the extent to which DNS "the protocol" (as opposed to DNS "the concept") is incapable of enforcing that everyo…

I'm not talking about just anchoring a hierarchy at an alternative root. What I mean is that as far as I know, nothing (DNSSEC notwithstanding) really prevents breaking away from the strict hierarchical model altogether and doing something like a system with web-of-trust or filtered by heuristics. From there it's possible to think about how to build a genuine consensus beyond "ICANN says these are the root servers; who am I to argue?". For TLS (DANE notwithstanding), I very well might misunderstand the situation but I thought it only mattered that the client, server, and CA agree on the server's name, not that they agree on a particular delegation of authority for assigning that name.

Re: Decentralized Web Primer

#58

Post by blunte »

da02 wrote: post #9
blunte wrote: post #2

Could be that I'm just really out of touch, but I need at least one paragraph saying "What is this?" (and ideally also, "Who is the target audience?" - or "Why would I be interested in reading this?")

I'm with you on this one. I had to Google all of this for a few minutes. This is my best interpretation below. I hope others will come in and correct my silly mistakes. (Disclaimer: I'm not a programmer and English is my 2nd language): "So you know how people use BitTorrent to distribute big-budget Hollywood movies with occasional mediocre storylines? What if you could distribute your own data on a similar P2P networ…

You're hired! Well done.

Re: Decentralized Web Primer

#59

Post by hedora »

bogomipz wrote: post #54
hedora wrote: post #50

Earlier quoted context omitted.

Those sites are all part of the decentralized web. Port 80 on my laser printer, router and flower pot are too. Using old well-defined terms to describe new concepts just leads to confusion. It would be like writing a story called “the history of the industrial revolution” that’s really about 3d printers.

So out of the one sentence that I wrote you decided to ignore the part where I wrote that decentralization was referring to centralization of content? Another word for centralization is consolidation. You might want to look up the term "walled garden": https://www.technologyreview.com/s/512316/facebook-and-googl...

I didn’t see the word “content” on first reading. I never have encountered Google or FB created content (other than open source code), and I also don’t use their news aggregators (or other services), so (with the exception of YouTube), I don’t think of them as being in the content game at all. Anyway, I’ve been happily ignoring their walled gardens for decades, and using the decentralized web for news, etc.

Certainly, they’re in the infrastructure + surveillance business, and (especially for Google) that’s concerning.

Also, “Decentralized” is a technical term. Google runs a large, decentralized infrastructure, based on web technologies.

“Consolidation” is usually used as a business term, and is orthogonal to centralizing operations. For instance, many companies have outsourced payroll to one centralized company. That does not mean their industries are consolidating.

When people intentionally use words wrong, it is bound to lead to confusion.

Re: Decentralized Web Primer

#60

Post by pessimizer »

jkarneges wrote: post #43
loup-vaillant wrote: post #23

Earlier quoted context omitted.

No. The web as it works right now is naturally centralised, for a simple reason: the bandwidth costs on a server is proportional to the size of its audience. This is why we need YouTube to begin with, instead of millions of people posting their videos on their personal page. There are other factors of course: firewalls, asymmetric bandwidth, security issues, technical ignorance… But the client-server model does play…

> This is why we need YouTube to begin with, instead of millions of people posting their videos on their personal page. People can post videos to their own hosted pages already. We "need" YouTube because 1) people generally don't want to host anything, and 2) it provides connectedness (search, related videos, subscribers, etc). P2P makes the first issue worse. The second issue is a problem of data federation, not cli…

> We "need" YouTube because 1) people generally don't want to host anything

No, because it is expensive to host things, as mentioned in the comment you are replying to.

Post reply on HN