Hi,
I saw this project and I thought it was very interesting:
Basically, it makes the clients connect to each other to share pages between each other using webrtc before going to the centralized server.
It would probably be a bad idea to convert mobile devices into network peers given the data restrictions and quality of connections but it seems like something very interesting for the desktop clients.
Cheers
It also sounds like a great way to expose your Wikipedia browsing to the cloud...
On Tue, Nov 22, 2016 at 5:01 PM Joaquin Oltra Hernandez < jhernandez@wikimedia.org> wrote:
Hi,
I saw this project and I thought it was very interesting:
Basically, it makes the clients connect to each other to share pages between each other using webrtc before going to the centralized server.
It would probably be a bad idea to convert mobile devices into network peers given the data restrictions and quality of connections but it seems like something very interesting for the desktop clients.
Cheers _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
See also related discussion last year https://lists.wikimedia.org/pipermail/wikitech-l/2015-November/084143.html
Personally I think this whole thing is a bad idea * Its questionable how much this would actually save anything. Cached anon hits are pretty cheap * This basically doesn't do cach invalidation. Lets just have vandalism stay around for long periods of time * Probably makes it much easier for third parties to determine what you are browsing. (Censorship resistant p2p networks is still an open research problem last I checked) * Probably makes it easier for adversaries to selectively censor specific articles [I haven't looked at the implementation, but I'm going to guess here] * Questionable how it would verify content is legit. What's stopping a malicious actor from putting random malicious js into the p2p network, or someone replacing articles with biased versions.
-- bawolff
On Tue, Nov 22, 2016 at 5:00 PM, Joaquin Oltra Hernandez jhernandez@wikimedia.org wrote:
Hi,
I saw this project and I thought it was very interesting:
Basically, it makes the clients connect to each other to share pages between each other using webrtc before going to the centralized server.
It would probably be a bad idea to convert mobile devices into network peers given the data restrictions and quality of connections but it seems like something very interesting for the desktop clients.
Cheers _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
For offline / poor connectivity use cases, I am more excited about https://wiki.mozilla.org/FlyWeb. In contrast to WebRTC based solutions, this enables completely local discovery and sharing of resources, without requiring an internet connection.
WebRTC based P2P CDNs are not very useful for the most common Wikipedia session, which is a single page lookup after following a link from a search engine. They are more useful for live video streaming, where session length, resource size, and number of simultaneous users interested in the same chunks is more favorable.
On Tue, Nov 22, 2016 at 11:33 AM, bawolff bawolff+wn@gmail.com wrote:
See also related discussion last year https://lists.wikimedia.org/pipermail/wikitech-l/2015-November/084143.html
Personally I think this whole thing is a bad idea
- Its questionable how much this would actually save anything. Cached
anon hits are pretty cheap
- This basically doesn't do cach invalidation. Lets just have
vandalism stay around for long periods of time
- Probably makes it much easier for third parties to determine what
you are browsing. (Censorship resistant p2p networks is still an open research problem last I checked)
- Probably makes it easier for adversaries to selectively censor
specific articles [I haven't looked at the implementation, but I'm going to guess here]
- Questionable how it would verify content is legit. What's stopping a
malicious actor from putting random malicious js into the p2p network, or someone replacing articles with biased versions.
-- bawolff
On Tue, Nov 22, 2016 at 5:00 PM, Joaquin Oltra Hernandez jhernandez@wikimedia.org wrote:
Hi,
I saw this project and I thought it was very interesting:
Basically, it makes the clients connect to each other to share pages between each other using webrtc before going to the centralized server.
It would probably be a bad idea to convert mobile devices into network peers given the data restrictions and quality of connections but it seems like something very interesting for the desktop clients.
Cheers _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
I'd love to have Wikimedia content not be dependent on WMF servers, but as noted by others, there are a lot of problems to address, including privacy and security issues. At some point I think it might be good to have an office hour or some other kind of meeting about this subject to discuss possibilities. I won't be involved in this for the for the foreseeable future due to the many other issues that are on my task list, but +1 moral support. I'd like to see reduced dependency of Wikimedia content on WMF included in the WMF strategic plan.
Pine
On Tue, Nov 22, 2016 at 11:48 AM, Gabriel Wicke gwicke@wikimedia.org wrote:
For offline / poor connectivity use cases, I am more excited about https://wiki.mozilla.org/FlyWeb. In contrast to WebRTC based solutions, this enables completely local discovery and sharing of resources, without requiring an internet connection.
WebRTC based P2P CDNs are not very useful for the most common Wikipedia session, which is a single page lookup after following a link from a search engine. They are more useful for live video streaming, where session length, resource size, and number of simultaneous users interested in the same chunks is more favorable.
On Tue, Nov 22, 2016 at 11:33 AM, bawolff bawolff+wn@gmail.com wrote:
See also related discussion last year https://lists.wikimedia.org/pipermail/wikitech-l/2015-
November/084143.html
Personally I think this whole thing is a bad idea
- Its questionable how much this would actually save anything. Cached
anon hits are pretty cheap
- This basically doesn't do cach invalidation. Lets just have
vandalism stay around for long periods of time
- Probably makes it much easier for third parties to determine what
you are browsing. (Censorship resistant p2p networks is still an open research problem last I checked)
- Probably makes it easier for adversaries to selectively censor
specific articles [I haven't looked at the implementation, but I'm going to guess here]
- Questionable how it would verify content is legit. What's stopping a
malicious actor from putting random malicious js into the p2p network, or someone replacing articles with biased versions.
-- bawolff
On Tue, Nov 22, 2016 at 5:00 PM, Joaquin Oltra Hernandez jhernandez@wikimedia.org wrote:
Hi,
I saw this project and I thought it was very interesting:
Basically, it makes the clients connect to each other to share pages between each other using webrtc before going to the centralized server.
It would probably be a bad idea to convert mobile devices into network peers given the data restrictions and quality of connections but it
seems
like something very interesting for the desktop clients.
Cheers _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
-- Gabriel Wicke Principal Engineer, Wikimedia Foundation _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
On Tue, Nov 22, 2016 at 11:33 AM bawolff bawolff+wn@gmail.com wrote:
See also related discussion last year https://lists.wikimedia.org/pipermail/wikitech-l/2015-November/084143.html
And the year before, and before that, and before that... it's a perennial proposal ;-)
Personally I think this whole thing is a bad idea
- Its questionable how much this would actually save anything. Cached
anon hits are pretty cheap
Indeed.
- This basically doesn't do cach invalidation. Lets just have
vandalism stay around for long periods of time
Yep, that's always been a problem in these proposals.
* Probably makes it much easier for third parties to determine what
you are browsing. (Censorship resistant p2p networks is still an open research problem last I checked)
Plus this.
- Probably makes it easier for adversaries to selectively censor
specific articles [I haven't looked at the implementation, but I'm going to guess here]
Basically because of the above.
- Questionable how it would verify content is legit. What's stopping a
malicious actor from putting random malicious js into the p2p network, or someone replacing articles with biased versions.
Same thing here. That being said....I'm curious if there's some sort of middle ground here. I wonder how much (c|w)ould be saved by serving static assets (CSS, UI images, etc etc) via P2P. Prolly not much in the US/Europe, but in places with poor latency this could be interesting.
-Chad
wikitech-l@lists.wikimedia.org