Post 9oBZXFns93UnsZdZKa by anna@witches.live
(DIR) More posts by anna@witches.live
(DIR) Post #9oB6zjyqdyVsuR0CsC by bortzmeyer@mastodon.gougere.fr
2019-10-22T08:37:39Z
0 likes, 0 repeats
A scientific paper about the fediverse: "Challenges in the Decentralised Web: The Mastodon Case" (executive summary: the #fediverse is not as decentralized as it claims) https://dl.acm.org/citation.cfm?id=3355572
(DIR) Post #9oB7WwxdP5xqivGoLI by Gargron@mastodon.social
2019-10-22T08:43:38Z
0 likes, 0 repeats
@bortzmeyer This... suggests something that already exists 😕 Moreover, it's kind of the most basic principle of Mastodon... That casts some doubt on how well-researched the paper is...
(DIR) Post #9oB8838TPPFoaa1wYa by bortzmeyer@mastodon.gougere.fr
2019-10-22T08:50:22Z
0 likes, 0 repeats
@Gargron I'm not sure. The content of text (not images) toots is stored in many places but the reference URL goes to one instance. Content may be in many places, but at URLs that people typically don't know.
(DIR) Post #9oB8Cgh8GsVaRBSAvQ by Gargron@mastodon.social
2019-10-22T08:50:15Z
0 likes, 0 repeats
@bortzmeyer Mastodon servers never subscribe to the entirety of another server, it's instance-to-user, not instance-to-instance
(DIR) Post #9oB8Ch2317ErU3EsWe by bortzmeyer@mastodon.gougere.fr
2019-10-22T08:51:13Z
0 likes, 0 repeats
@Gargron I'm not the author(s), you should report to them :-)
(DIR) Post #9oB8F8HWfsLEZmds7E by Gargron@mastodon.social
2019-10-22T08:51:39Z
0 likes, 0 repeats
@bortzmeyer Yeah but the content remains available in followers' feeds throughout outages on the origin. The wording suggests they believe content is pulled whenever people load their home feeds..?
(DIR) Post #9oB8iIgX4ectEuFN6e by Gargron@mastodon.social
2019-10-22T08:56:54Z
0 likes, 0 repeats
@bortzmeyer I am trying to provide context for you and those looking at your post.
(DIR) Post #9oB9EE41n4J8ArSZyy by bortzmeyer@mastodon.gougere.fr
2019-10-22T09:02:41Z
0 likes, 0 repeats
@Gargron If you still have a Twitter account, you can discuss with @_aravindh@twitter.com.
(DIR) Post #9oB9UqmHcZiQsO5Sxk by lain@pleroma.soykaf.com
2019-10-22T09:05:39.592104Z
1 likes, 0 repeats
@Gargron @bortzmeyer 5 authors and they still didn't get how things work, nice.
(DIR) Post #9oB9ngCXAX7nZkLPOq by bortzmeyer@mastodon.gougere.fr
2019-10-22T09:09:06Z
0 likes, 0 repeats
@lain @Gargron Careful not to focus on some clumsiness in their paper, and forget the real issues they raise. (In other words: let's not be defensive.)
(DIR) Post #9oBA6vEvTotjiddfG4 by lain@pleroma.soykaf.com
2019-10-22T09:12:33.024540Z
2 likes, 0 repeats
@bortzmeyer @Gargron i read the paper a while ago, i did not find any new issues raised that people were not aware of. I guess we're not the audience for a paper like this, anyway.
(DIR) Post #9oBAGsiLiy2omZ9NHU by 1iceloops123@shitposter.club
2019-10-22T09:14:27.887091Z
0 likes, 0 repeats
@lainIf anyone here could write the history of the fedi maybe we would have a accurate paper or book. Like I'm to dumb@bortzmeyer @Gargron
(DIR) Post #9oBAe6pvNtj7F3ll0i by bortzmeyer@mastodon.gougere.fr
2019-10-22T09:18:34Z
0 likes, 0 repeats
@lain @Gargron "aware of" is one thing, acting on it is another. My personal unscientific feeling is that many instance admins ignore the inconvenient problems (such as the certificate issues mentioned in the paper).
(DIR) Post #9oBE2R524Px2q57wxs by Gargron@mastodon.social
2019-10-22T09:56:28Z
0 likes, 0 repeats
@bortzmeyer @lain I mean, you by definition expand the pool of who can host a social media server from 200-engineer corporations to the common person. I find it neither surprising nor damning that a large number of nodes are not run professionally. What's important is that they can be run, and that some of them are run professionally.
(DIR) Post #9oBFzA1dDtgez46QLI by schmittlauch@toot.matereal.eu
2019-10-22T10:18:23Z
1 likes, 0 repeats
@bortzmeyer @lain @Gargron The issue they raise should be well-known throughout the fediverse, but the solution they propose (disclaimer: only according to these screenshots) is not fit for purpose due to their massive misunderstanding of data flows.I'm more looking forwards to @cwebber's ideas of content-adressable distribution using datashards, DIDs and stuff.
(DIR) Post #9oBYZxwctsHdPDSecS by theruran@hackers.town
2019-10-22T13:46:41Z
0 likes, 0 repeats
@bortzmeyer did they measure it or just wave their hands?
(DIR) Post #9oBYoFC9YnYSxeX3Bo by bortzmeyer@mastodon.gougere.fr
2019-10-22T13:49:19Z
0 likes, 0 repeats
@theruran Another ultra-defensive toot. Mastodon people do not like criticisms, it seems.(Read the paper, it is very detailed.)
(DIR) Post #9oBYvkc791cGQpFqiG by Gargron@mastodon.social
2019-10-22T08:55:29Z
0 likes, 0 repeats
@bortzmeyer And oh, so this is what King's College IPs were doing a while ago. It was a notable point of discussion among Mastodon admins at the time 😬
(DIR) Post #9oBYvlBuzuFwDs0RWK by anna@witches.live
2019-10-22T13:42:33Z
0 likes, 0 repeats
@Gargron @bortzmeyer all this scraping of mastodon instances, and they couldn't even be bothered to learn how federation works smdh
(DIR) Post #9oBYvm0xw75amBtD5U by bortzmeyer@mastodon.gougere.fr
2019-10-22T13:50:40Z
0 likes, 0 repeats
@anna @Gargron Wow, a lot of arrogance from Mastodon people, in response to this article. There is very little documentation on the internal of the federation (the ActivityPub standard is just a small part) and, when outsiders start to look at it, the community reacts quite violently.
(DIR) Post #9oBZ80etJfPrfWOnYW by theruran@hackers.town
2019-10-22T13:52:53Z
0 likes, 0 repeats
@bortzmeyer just wondering if it's a good paper or not
(DIR) Post #9oBZDLF8D8U8xTtOrI by bortzmeyer@mastodon.gougere.fr
2019-10-22T13:53:52Z
0 likes, 0 repeats
@theruran IMHO, it is (which does not mean it is above criticism).
(DIR) Post #9oBZUigKJtAoZhtpGy by radikalgrafitio@sunbeam.city
2019-10-22T13:56:58Z
0 likes, 0 repeats
@bortzmeyer @anna @Gargron perhaps if they had come and asked?
(DIR) Post #9oBZXFns93UnsZdZKa by anna@witches.live
2019-10-22T13:57:25Z
0 likes, 0 repeats
@bortzmeyer @Gargron scraping our instances is both expensive to us and invasive of our community. if someone is doing a research paper they should research how the software works before making assumptions. if someone can write a scraper, they can easily observe from an existing instance, or by setting up their own, that instance servers keep a copy of remote statuses (and media) and serve it to users so that remote servers being small or down don't negatively effect the experience
(DIR) Post #9oBaCc4AHG0onaymhM by nutsling@mastodon.social
2019-10-22T14:04:55Z
0 likes, 0 repeats
@bortzmeyer @anna @Gargron one critical toot = violence
(DIR) Post #9oBhrofl0lt5Rt9oe0 by bortzmeyer@mastodon.gougere.fr
2019-10-22T15:30:49Z
0 likes, 0 repeats
@nutsling @anna @Gargron Saying that the researchers "couldn't even be bothered to learn how federation works"' is not "a critical toot'", it's insulting.
(DIR) Post #9oBixbSHANS89CAtea by nutsling@mastodon.social
2019-10-22T15:43:04Z
0 likes, 0 repeats
@bortzmeyer @anna @Gargron its probably more insulting to researchers who take their field seriously rather than those who only get involved to further their politics, welp see ya later
(DIR) Post #9oBkU44iORjW8DCHaK by esvrld@octodon.social
2019-10-22T16:00:05Z
0 likes, 0 repeats
@bortzmeyer @nutsling @anna @gargron if i were doing research on the fediverse, i would simply, research how federation works first, therefore eliminating the issue
(DIR) Post #9oDKpJ0DsZMA19U2fA by imcpaperauthors@mastodon.social
2019-10-23T09:44:41Z
0 likes, 0 repeats
@Gargron @bortzmeyer @aravin Hey all, we’re the authors on the paper under discussion. Thanks for the comments! We are really glad to see it being critiqued by core users and developers of Mastodon. Hope nobody felt the paper was an "attack" on the concept of Mastodon - we're all big fans, and we would really like for it to succeed - the paper was meant as a first step to quantify challenges (hopefully in a useful way). 1/4
(DIR) Post #9oDKpJVQ0aJHZu4xHs by imcpaperauthors@mastodon.social
2019-10-23T09:47:49Z
0 likes, 0 repeats
@Gargron @bortzmeyer @aravin We thought it might be useful to try and address a couple of the points raised. As @Gargron mentioned, the replication evaluation wasn't about ensuring access for existing followers. It's more about retaining universal access to toots (e.g. even when a user has no followers). 2/4
(DIR) Post #9oDKpK7hiEw1UdzWxk by imcpaperauthors@mastodon.social
2019-10-23T09:48:10Z
0 likes, 0 repeats
@Gargron @bortzmeyer @aravin The other point about the instance-to-instance subscription is a bit more complex, but we didn't mean to imply that instances subscribe to all posts on other instances - the graph was induced from the follower network. Sorry if our language was imprecise here: we essentially mean to say that if user_a@instance1 subscribes to user_b@instance2, then we induce an edge from instance1 to instance2, as toots from instance2 will need to get replicated on instance1. 3/4
(DIR) Post #9oDKpKeJkz1T7nFZnU by imcpaperauthors@mastodon.social
2019-10-23T09:48:35Z
0 likes, 0 repeats
@Gargron @bortzmeyer @aravinPlease feel free to fire more critique/questions :) If useful, we could even spin up a blog to answer some of your concerns. We are also happy to contribute in other ways the Mastodon community might find useful (e.g. trying to convert some of the ideas into code in collaboration with the community). 4/4
(DIR) Post #9oDKpL8nvdPQeLVvJg by Gargron@mastodon.social
2019-10-23T10:01:34Z
0 likes, 0 repeats
@imcpaperauthors @bortzmeyer @aravin I don't have any substantial critiques after these points have been clarified. I disagree in the value judgement somewhat, in particular:Given how easy it is to spin up a Mastodon server, it should not be surprising that there is a high churn in low-user servers. Some are just trying it out. Similarly, a lot of servers are personal (i.e. for one person only). In that case downtimes don't affect anyone but that person.
(DIR) Post #9oDKpLaoFVoK3CcHy4 by Gargron@mastodon.social
2019-10-23T10:05:50Z
0 likes, 0 repeats
@imcpaperauthors @bortzmeyer @aravin Likewise, very rarely is there a case where someone wants to access a post from some external source that isn't a home feed / federated timeline / cached profile. Most activity happens between followers. So public unavailablity isn't as catastrophic as the paper would make one believe.
(DIR) Post #9oDKpLyunt5pFxtXXc by bortzmeyer@mastodon.gougere.fr
2019-10-23T10:21:58Z
0 likes, 0 repeats
@Gargron @imcpaperauthors @aravin I disagree here. There are different usages. One of the tenets of ActivityPub being that "everything has an URL and is retrievable", some people use fediverse URLs in blog posts, Twitter tweets, email, etc. It seems legitimate to me.
(DIR) Post #9oHOqazle3z8YjysfA by ella_kane@hackers.town
2019-10-25T09:25:57Z
0 likes, 0 repeats
@bortzmeyer fedi or masto? I'm curious anyway. Must read. Thanks!
(DIR) Post #9oHP8qXCbpkZxxK3Ie by bortzmeyer@mastodon.gougere.fr
2019-10-25T09:29:16Z
0 likes, 0 repeats
@ella_kane The title, and most of the paper, mentions Mastodon, but actually, the issues raised are common to all the fediverse.
(DIR) Post #APUcz4JHm0XhxLRiSm by ignactro@mastodon.social
2022-11-11T11:43:18Z
0 likes, 0 repeats
@bortzmeyer we keep doing some work trying to understand challenges in the DW (https://arxiv.org/pdf/2204.12709 & https://arxiv.org/pdf/2110.13500) and also at the IETF if you wanna grab a coffee :)