10 ms·

Removing the v2 code from clients and falsely stating v2 onions aren't valid URLs is lying to users to protect them from a potential future threat. But that pro

Removing the v2 code from clients and falsely stating v2 onions aren't valid URLs is lying to users to protect them from a potential future threat. But that protection comes at the cost: the destruction of the entire 15 year history of communities and interlinking and search indices for onionland. To protect the community they have destroyed it. A v3 community might come in the future, and I hope it will, but right now v3 .onions are pretty much just bots and a few big names that have created new v3 services. The rest is botnet v3 .onions. What I want, what everyone in the tor project comment thread on the depreciation post wants, is for v2+v3. Not just v2. Even now most tor relays support v2 and will until the tor project puts out a version consensus flag that blocks older tor. But they can't do that any time soon because most tor infrastructure still supports v2.

ziddoap 5y ago
I'm still not understanding why the 15 year history of communities was not able or willing to act on the 13 months of warning, and months of discussion before that. There was more than ample time to prepare. Surely you don't expect a project like Tor to simply ignore security forever going forward because "people still use it"? Should Tor just stay V2, security be damned? I don't think Tor's main goal is to ensure your community stays friends; the goals are a bit loftier. I wonder if you hold the same views on any other security-related thing like cryptography or password storage. We'd still be at 8 character, lower-case only, truncated passwords "because people built infrastructure for it".
- npteljes 5y ago
  I just don't get GP's point. If the workings of the system compromise its entire premise, of course they upgrade the system so that it matches the original intentions. It's like they shouldn't have deprecated old SSL and TLS versions, because some web services still only use those. I can't imagine an active community having trouble with this.
  - saurik 5y ago
    The problem is that the URLs themselves got deprecated: the entire interlinked ecosystem got destroyed. Maybe think of what would happen if all of our current TLDs got deprecated tomorrow, and in a year or two support for them would start getting dismantled. Every single link anyone has ever posted on Hacker News is now broken. Now what? You are all making it sound like this isn't some big deal: that people just needed to "get their act together"... but how do you rebuild the world wide web after breaking every link in the web? Whose job is it even to fix every single link on every single page and stored in every single comment in every single database? Now, you can certainly argue that the old system was broken by insecurity and thereby had to die. But that just means the entire concept of onion services was some temporary art project more than a platform on which people could build the kind of ecosystem we have on the "normal" world wide web. At best, it means that people needed to come up with some extremely different way of thinking about what those .onion addresses really were, as they were clearly more like IP addresses than hostnames. I personally can't help but think there could have been some better transition plan. Like maybe a mechanism where people can leave cross-signed redirects from v2 to v3 domains, and maybe even they would only be honored if signed before some date, or maybe there is a disambiguation dialog for competing redirects that also shows the time the redirect was published (these being attempts to deal with "in the future someone could crack this and start making arbitrary redirects for any old site"). But that clearly didn't happen and now that entire ecosystem is dead. Maybe it will rebuild, and maybe one day it will be stronger than it is now, but frankly... when that happens, wouldn't you expect people to start pointing out that the cryptography of the ECC curve they are using is now subject to breaking by quantum computers or whatever we have in another 15 years? Maybe they will handle that transition better, or maybe the mere thought of that will serve as a lesson to not rebuild inside that design. Truly, though, a design that doesn't take this into consideration from the beginning is inherently flawed. As I noted in passing earlier: these .onion names were clearly more like IP addresses, and so you needed some kind of translation later on top of that to provide later portability. As much as I often hate on the current players in "decentralized DNS" (for various reasons that they could have avoided), that's the kind of concept that clearly is going to win the day going forward.
    - ziddoap 5y ago
- superkuh 5y ago
  Because tor v2 and v3 are not compatible. You can't just make all the links everyone ever made on any site, or entries in any search index, suddenly point to a new URL. You can announce v2 is going to disappear and say that everyone should try to port what they can, but most don't. Only the big sites and their users actually have any pseudo-continuity through their efforts. And even then all the links stop working.
  - ziddoap 5y ago
    The fact that people didn't port is not an argument that they shouldn't have ported. Tor is primarily a security and privacy project. Arguing that they should keep an insecure version because people decided to ignore the security implications of that insecure version is an awful argument. Especially when those people had well over a year to execute. Your use case for Tor seems to be unrelated to the projects goals.

ziddoap 5y ago

I'm still not understanding why the 15 year history of communities was not able or willing to act on the 13 months of warning, and months of discussion before that. There was more than ample time to prepare. Surely you don't expect a project like Tor to simply ignore security forever going forward because "people still use it"? Should Tor just stay V2, security be damned? I don't think Tor's main goal is to ensure your community stays friends; the goals are a bit loftier. I wonder if you hold the same views on any other security-related thing like cryptography or password storage. We'd still be at 8 character, lower-case only, truncated passwords "because people built infrastructure for it".

npteljes 5y ago
I just don't get GP's point. If the workings of the system compromise its entire premise, of course they upgrade the system so that it matches the original intentions. It's like they shouldn't have deprecated old SSL and TLS versions, because some web services still only use those. I can't imagine an active community having trouble with this.
- saurik 5y ago
  The problem is that the URLs themselves got deprecated: the entire interlinked ecosystem got destroyed. Maybe think of what would happen if all of our current TLDs got deprecated tomorrow, and in a year or two support for them would start getting dismantled. Every single link anyone has ever posted on Hacker News is now broken. Now what? You are all making it sound like this isn't some big deal: that people just needed to "get their act together"... but how do you rebuild the world wide web after breaking every link in the web? Whose job is it even to fix every single link on every single page and stored in every single comment in every single database? Now, you can certainly argue that the old system was broken by insecurity and thereby had to die. But that just means the entire concept of onion services was some temporary art project more than a platform on which people could build the kind of ecosystem we have on the "normal" world wide web. At best, it means that people needed to come up with some extremely different way of thinking about what those .onion addresses really were, as they were clearly more like IP addresses than hostnames. I personally can't help but think there could have been some better transition plan. Like maybe a mechanism where people can leave cross-signed redirects from v2 to v3 domains, and maybe even they would only be honored if signed before some date, or maybe there is a disambiguation dialog for competing redirects that also shows the time the redirect was published (these being attempts to deal with "in the future someone could crack this and start making arbitrary redirects for any old site"). But that clearly didn't happen and now that entire ecosystem is dead. Maybe it will rebuild, and maybe one day it will be stronger than it is now, but frankly... when that happens, wouldn't you expect people to start pointing out that the cryptography of the ECC curve they are using is now subject to breaking by quantum computers or whatever we have in another 15 years? Maybe they will handle that transition better, or maybe the mere thought of that will serve as a lesson to not rebuild inside that design. Truly, though, a design that doesn't take this into consideration from the beginning is inherently flawed. As I noted in passing earlier: these .onion names were clearly more like IP addresses, and so you needed some kind of translation later on top of that to provide later portability. As much as I often hate on the current players in "decentralized DNS" (for various reasons that they could have avoided), that's the kind of concept that clearly is going to win the day going forward.
  - ziddoap 5y ago
superkuh 5y ago
Because tor v2 and v3 are not compatible. You can't just make all the links everyone ever made on any site, or entries in any search index, suddenly point to a new URL. You can announce v2 is going to disappear and say that everyone should try to port what they can, but most don't. Only the big sites and their users actually have any pseudo-continuity through their efforts. And even then all the links stop working.
- ziddoap 5y ago
  The fact that people didn't port is not an argument that they shouldn't have ported. Tor is primarily a security and privacy project. Arguing that they should keep an insecure version because people decided to ignore the security implications of that insecure version is an awful argument. Especially when those people had well over a year to execute. Your use case for Tor seems to be unrelated to the projects goals.