[{"author": "Christian Huitema", "text": "<p>where is the chat room for dinrg? If I click on the agenda, I get the chat room for the last meeting in July!</p>", "time": "2023-11-06T09:20:44Z"}, {"author": "Dirk Kutscher", "text": "<p>Ouch, sorry about that. Yes, this is the chat room.</p>", "time": "2023-11-06T09:21:18Z"}, {"author": "Christian Huitema", "text": "<p>I see it now!</p>", "time": "2023-11-06T09:21:59Z"}, {"author": "Dirk Kutscher", "text": "<p>I closed the queue after Christian.</p>", "time": "2023-11-06T09:23:56Z"}, {"author": "Lixia Zhang", "text": "<p>@Christian: yes it is natural to have DHT nodes in the cloud.It is just that this fact is againt the claim of \"decentralization\" of not being controlled by cloud providers.</p>", "time": "2023-11-06T09:31:41Z"}, {"author": "Christian Huitema", "text": "<p>I am not sure that we <em>need</em> to have the DHT nodes in the cloud. Some DHT work with long duration connection between nodes in their routing tables, and that could use NAT traversal. But you probably need a smattering of cloud services to seed the NAT traversal.</p>", "time": "2023-11-06T09:36:41Z"}, {"author": "Christian Huitema", "text": "<p>I meant cloud nodes, not services. It's 1:40am here...</p>", "time": "2023-11-06T09:40:04Z"}, {"author": "Lixia Zhang", "text": "<p>The problem of NOT using cloud nodes: see bullet #4 on take away slides: Non-cloud hosts come and go, causing issues with reliability</p>", "time": "2023-11-06T09:42:56Z"}, {"author": "Christian Huitema", "text": "<p>Do we have stats on this \"instability\"? Respective share of nodes going to sleep, losing connectivity, rebooting, doing software updates, etc? If I was doing that, I would put the DHT on a dedicated Ras.Pi, and have the other local nodes access that...</p>", "time": "2023-11-06T09:46:44Z"}, {"author": "David Guzman", "text": "<p>It really depends on the use case which is run over the DHT, for the case of IPFS it seems that it goes to hours [Trautwein 2022] (Design and evaluation of ipfs: A storage layer for the decentralizedweb)</p>", "time": "2023-11-06T09:56:33Z"}, {"author": "David Guzman", "text": "<p>(Fig. 8)</p>", "time": "2023-11-06T09:56:51Z"}, {"author": "Tianyuan Yu", "text": "<p>A quick question for the presenter: Without cloud providing identities, how the local-first software authenticate individual users (which I think is needed for end-to-end encryption)?</p>", "time": "2023-11-06T10:01:17Z"}, {"author": "Micha\u0142 Kr\u00f3l", "text": "<p>@Christian Huitema: there are quite a lot of stats here: <a href=\"https://probelab.io/ipfsdht/\">https://probelab.io/ipfsdht/</a></p>\n<p>Hosting a DHT on a Pi would indeed work but I'm not sure how many regular users will be able (and willing) to do so.</p>", "time": "2023-11-06T10:03:31Z"}, {"author": "Christian Huitema", "text": "<p>This is the kind of problems that get fixed once enough people focus on them. Maybe consider categories of nodes, put structure in the DHT, etc. Not that I know what to do, but it seems doable.</p>", "time": "2023-11-06T10:12:51Z"}, {"author": "Lixia Zhang", "text": "<p>Follow Michal comment about able or willing: in particular if one looks into some of use cases of IPFS, e.g. see 2 cases captures at <a href=\"https://en.wikipedia.org/wiki/InterPlanetary_File_System\">https://en.wikipedia.org/wiki/InterPlanetary_File_System</a></p>", "time": "2023-11-06T10:13:08Z"}, {"author": "Lixia Zhang", "text": "<p>@Geoff Huston: I agree with tech disruptions can disturb (incumbents, but disagree (how dare am I:)  that tech advances are represent only by Moore's law.</p>", "time": "2023-11-06T10:31:05Z"}]