Hacker News Reader

Comment by jameslk | Hacker News Reader

PeerTube exposes the IPs of all viewers who watch videos[0]. This is a privacy issue that's baked into its design. For example, this allows for anyone to track the content viewing history of IPs. I see this being a big hurdle for any mainstream adoption.

0. https://github.com/Chocobozzz/PeerTube/issues/316

View on HN

Without this it's hard to imagine PeerTube being able to scale to a fraction of YouTube's level. Video streaming is just too expensive. Being able to offload it to viewers during high traffic is a killer feature.

Google handles this in the same way as Netflix by living in your ISP's data centers. Google has an edge global cache that can live at ISPs to reduce bandwidth usage.

https://peering.google.com/#/options/google-global-cache

Google and Netflix can do this, because ISPs have to let them host there to avoid massive burden on their network. An individual who wants to host their own videos cannot. So, as Sir_Cmpwn indicates, some sort of torrent-esque system is needed to let individual users have some sort of option that avoids crazy bandwidth costs.

Does torrent or any other protocol out there try to prioritize peers that are within the same ASN or something similar? Many protocols ignore the physical network, but here it seems like it would help to be smarter.

Huh, this is a really good question/idea.

There's no need for manual priorization for nearer ASN's on BitTorrent, as is this builtin implicitly by prioritizing faster connections with higher bandwidths automatically. If a faster connection exists in a different ASN, the faster one is prioritized. Nuclear proof dynamic design, similar to TCP/IP.

I heard that BitTorrent is 40% traffic of today Internet

I've been digging around peer tube for a few days - a lot of cool stuff there.

I wonder if the barriers will largely be the costs of using it relative to existing centralized systems:

- uptime/accessibility over time of content : what how do users find content bookmarked two years ago if nodes are coming and going. Assuming there is a way to keep an updated bookmark list, what are the costs (effort, emotional etc) to the user to do that?

- costs of maintaining a node in the decentralized system. if my node maintains a catalog of 1000 or more videos what will be the operational costs (not just the storage and compute costs) to me of keeping this running.

These things are great to think of in context of of the recent HN post that discusses decentralized systems https://news.ycombinator.com/item?id=17361199

PS. I'm not saying that it's impossible to overcome those points - just that they're difficult and worth working through.

People pay for seedboxes all the time. If libtorrent adds webtorrent support I can see these seedboxes being repurposed to host webtorrent videos. As a seedbox owner you would just add your favorite channel's rss feed to the auto download.

thanks for that - I'm a little embarrassed to say I didn't even know what a seedbox was until now.

It uses BitTorrent to scale, so it must do so. This is so that your instance wouldn't go down if your video blows up. I consider this a tradeoff worth making, do you have a better solution?

Seed torrents you haven't watched to obscure what you are actually watching.

> this allows for anyone to track the content viewing history of IPs

Is that worse than only Google being able to do it?

Also tracking the viewing history of an IP is not that easy. People trying to do this can be easily spotted by the trackers.

Considering that Google is a subset of anyone, I have to assume that it is.

If you consider Google being able to track something as a negative, then Google+n would be at last +n negative.

There's an up-and-coming decentralized project called Force Network that has a method to obfuscate the ip of both the consumer and the host, and a mechanism to make it extremely censorship-resistant.

From the lightpaper presented, it seems to me, it's just tor with cryptocurrency. Am I misreading it and would you be so kind as to provide an executive summary why would one choose Force over good ole Tor?

See also https://github.com/Chocobozzz/PeerTube/issues/685 for discussion on implementing webtorrent over i2p

Technically it's to provide an opt-out switch in the interface. It's being developped and will land in the stable.

This shouldn't be an opt out, it should be the default. Most won't understand the ramifications of not opting out.

Breaks the distribution and cost model. You either need to join the swarm or pay to watch the video from a CDN. There is a cost to deliver bits, it's just been hidden from you by corporations subsidizing it. Therefore, it can't be default (although a modal could be provided that informed you prior to streaming and gave you the chance to bail out).

They used to make you pay extra for a static IP, but Comcast basically gives me a static IP for free now... I have to spoof my router's MAC address to get a new IP... I guess if you do that weekly, it is not such a huge privacy issue...

We could discuss the possible workarounds for this, but let's not forget that in terms of privacy, Youtube has its own by-design flaws as well. The company gathers much precise data about viewers than Peertube does.

Exposing your IP address to the public is much more risky than exposing it to Google.

I don’t think that’s a verbatim claim you can just make and expect everyone to agree with without any form of argument.

Counter-argument one: things known by Google represents another dose of data into a single place which accumulates way too much of it already.

If I watch something on a peertube, many may know that my IP streamed that, but they don’t know who I am because they’re not an all knowing internet-scale privacy-violator. Best of all, google won’t know I watched it at all, so it won’t be another annoying data-point in the super-aggressive youtube bubble.

> Best of all, google won’t know I watched it at all

Why do you think that? It would be very easy for Google to observe who is viewing PeerTube videos and to link that back to those people's YouTube profiles. You may trust them not to, but if you do, it seems you're half way to thinking it's OK for them to know.

> It would be very easy for Google to observe who is viewing PeerTube videos

I'm not saying it can't be done, but I wouldn't assume they would go out of their way to monitor this, just like I wouldn't assume they're monitoring public IPs of torrents in the process of downloading.

Yes. It would be easy. But why would they?

Speaking in BT-terms... Peertube's might find this peeking annoying and start publishing "peerguardian" like lists to prevent Google-spying.

Well it's kind of like driving a car with plates. It can identify you, but there's only so much risk.

You can also deactivate WebRTC and get a direct HTTP streaming fallback.

Would a VPN help in that regard?

Sure, or Tor. Anything that hides your IP.

Although IIRC there are gotchas specifically around webrtc sometimes bypassing VPNs if not configured exactly right.

I don't care. Amd people can VPN.

Reply on news.ycombinator.com