Notices by MMN-o ✅⃠ (mmn@social.umeahackerspace.se), page 4
-
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Thursday, 27-Jul-2017 15:28:51 UTC MMN-o ✅⃠ #WhatsApp fetches websites linked in the chat from your phone and thus makes OTR/encryption totally pointless. -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Wednesday, 19-Jul-2017 16:13:45 UTC MMN-o ✅⃠ I'd like to interject for a moment and mention that it is in fact Free Software, not just Open Source. This means you have the right not only to "view source" but also to build upon and the obligation to redistribute these modifications to the community so we together build the commons that makes the world stay afloat. -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Saturday, 15-Jul-2017 12:51:42 UTC MMN-o ✅⃠ @rugk The method with which updates are delivered in !GNUsocial (to subscribers). -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Friday, 14-Jul-2017 20:49:22 UTC MMN-o ✅⃠ Btw @chimo could I ask of you (and anyone else reading this too!) to do the https://websub.rocks tests for !gnusocial and submit the results as per instructions on the site? -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Wednesday, 12-Jul-2017 13:53:04 UTC MMN-o ✅⃠ @hikerus Mastodon had in a previous version (probably still widely in use?) an antifeature that ended push-subscriptions with a slightly too trigger-happy assumption of when a remote server has disappeared.
While !GNUsocial happily continues several months after initial failures .) -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Tuesday, 11-Jul-2017 19:07:49 UTC MMN-o ✅⃠ Ooh, yes. Make the #WebMention (LinkBack) plugin work properly in !GNUsocial. _THAT_ is my vacation project. Oh how I've wanted to do this. I will make @kevinmarks proud. -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Tuesday, 11-Jul-2017 10:53:33 UTC MMN-o ✅⃠ In shocking news, delivered to you via a commit message for !GNUsocial: Noone uses Facebook anymore.
https://git.gnu.io/gnu/gnu-social/commit/e4d77cb9b29efee5c4baa535c3a71d99610d4359 -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Tuesday, 11-Jul-2017 10:26:17 UTC MMN-o ✅⃠ Here's to breaking !GNUsocial instances: ☕ -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Monday, 10-Jul-2017 19:00:02 UTC MMN-o ✅⃠ @headcrack !GNUsocial stalls because the underlying HTTPS request never times out, which only happens with PHP sockets as the backend for HTTP_Request2 and not with the CURL backend. This is reproducible outside of the GNUsocial framework as well: https://git.gnu.io/gnu/gnu-social/issues/281#note_5674
I am fully aware of the snail-pace type DoS attack and that's why there is a "full spectrum timeout" set in deeper parts of the code (HTTP_Request2_SocketWrapper runs stream_set_timeout - and PHP itself has a default_socket_timeout of 60s), which if the socket timeouts worked (in PHP) would kill the connection _regardless_ of how many or few bytes have been received since the last fread(): -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Monday, 10-Jul-2017 18:53:49 UTC MMN-o ✅⃠ @sen @headcrack Unless I'm horribly mistaken, that's an entirely separate timeout. I.e. for the daemon->subprocess. The subprocess on its own has its entirely own timeouts for each individual instruction it runs.
Also, if this is supposed to relate to the hash.my ordeal, it doesn't explain why CURL as a backend handles it perfectly fine while PHP sockets don't.
Also I have run tests entirely separate of the !GNUsocial framework and can recreate the issue of inconsistent timeouts with PHP sockets: https://git.gnu.io/gnu/gnu-social/issues/281#note_5674 -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Monday, 10-Jul-2017 13:01:57 UTC MMN-o ✅⃠ @shnoulle 60 is the original default. Nothing should ever take more than 60 seconds to complete for a webserver that remote users can initiate requests for (i.e. link stuff that gets looked up by your !GNUsocial server). If it was up to me I'd set it to 30 seconds or something.
oEmbed etc, which in combination with StoreRemoteMedia does remote downloads, has an upper limit anyway on how large files to download (so 30 seconds should be reasonable unless your server is on a really slow connection). -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Monday, 10-Jul-2017 12:52:56 UTC MMN-o ✅⃠ I like the cap on !GNUsocial. -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Monday, 10-Jul-2017 12:29:51 UTC MMN-o ✅⃠ Hey everybody, !GNUsocial seems to be working fine with #PHP 7.2alpha3 after just a few patches!
Also found a bug in !qvitter thanks to updating to 7.2! https://git.gnu.io/h2p/Qvitter/merge_requests/102 -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Sunday, 09-Jul-2017 17:58:50 UTC MMN-o ✅⃠ @shnoulle Well there is a configurable option for timeout (not connection timeout, i.e. establishing it) but it doesn't seem to be enabled by default in the HTTP library: 'timeout' => 0,
I know I've been fiddling with this at some time and knowing full well that there needs to be a timeout for the whole connection (a famous, now mostly taken care of, httpd DoS attack is to just reeeeaaaalllyyy sloooowwwlllyyyy byte by byte transfer a simple HTTP request, which can easily cause a zillion open connections).
I'll have a look if there's something set somewhere that I've missed so far. (I enjoy having !GNUsocial debugging sessions in the !fediverse btw .]) -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Sunday, 09-Jul-2017 09:23:20 UTC MMN-o ✅⃠ @Gargron The data in !OStatus travels with HTML representation. If you create the HTML representation in a way that puts <img/> tags where the image links are (you know, like blog posts) then it's shown correctly.
...except that microblogging sites like !GNUsocial tend to ignore <img/> tags because we haven't bothered matching them against the attachments and presenting them locally to avoid third party tracking...
My point is OStatus is perfectly fine to represent interspersed text and images (because it's HTML) and it's only up to the client to generate good HTML and the subscriber to trust it. Just use some sort of WYSIWYG, like Wordpress, to post your stuff and let PuSH/WebSub distribute it for you (or implement WYSIWYG editing in Mastodon for example).
cc: @waha06x36 @halcy -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Sunday, 09-Jul-2017 08:46:58 UTC MMN-o ✅⃠ Yeah I have noticed twice that this happens too. Didn't see anything at a quick glance yesterday evening but will look harder today. Obviously something related to !GNUsocial daemons. -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Monday, 03-Jul-2017 06:45:40 UTC MMN-o ✅⃠ @gargron preferrably something at least compatible with Wikipedia, say CC:by-sa. We use CC:by in !GNUsocial -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Monday, 03-Jul-2017 06:43:31 UTC MMN-o ✅⃠ @Gargron I really think you should impose a default license to #Mastodon creations. A libre CC license supposedly. Otherwise copyright says noone is allowed to redistribute neither posts nor other media. Formality may suck but it would suck more if ysers start harassing other instances with takedown requests because they didn't understand that posting on the web means copying and distributing by nature. -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Thursday, 29-Jun-2017 07:33:23 UTC MMN-o ✅⃠ @mikegerwitz Oh, isn't it clickable? That might be a regression on my part. Good !GNUsocial bug report! -
MMN-o ✅⃠ (mmn@social.umeahackerspace.se)'s status on Sunday, 25-Jun-2017 08:10:00 UTC MMN-o ✅⃠ I think we should do whatever !xmpp does with usernames. There are many lessons to be learned there (unicode treachery etc.)
Also, yes I think URLs, as it is in !GNUsocial already, should be possible to mention. Not just acct: style URIs.
I think it's important to remember that some users may want just their domain name as identifier for example, which btw makes parsing/prioritising a bit trickier if usernames can have dots in them ;) But I think there are solutions.
And on that topic, usernames can be @ but don't support '.' in GS. If supported we'd probably prefer the local username if not appended by '/'.