@kas@enotty.dk that sounds fair to me.
Additionally, there’s a lot that can be done by a client to reduce the network traffic and UI latency of twtxt without changing the protocol.
@mdom@domgoergen.com, @kas@enotty.dk re: metadata, I’m (obviously) in favor of my suggestion for metadata-in-comments, but I don’t think we should have comments in comments.
@mdom@domgoergen.com, @dave@davebucklin.com, @kas@enotty.dk I agree with the “no max length; show at least 140chars” idea.
@phil@philmcclure.duckdns.org I don’t just throw away lines i can’t parse, but try to fix as many errors as possible. But eventually you have to give up. https://github.com/mdom/txtnish/blob/master/bin/txtnish#L368
@phil@philmcclure.duckdns.org I can’t imagine a simpler time format than rfc3339. I would be fine with just using utc or unix time, but i don’t think most people could write that without a client… :)
@dave@davebucklin.com, @kas@enotty.dk I thought about propsing “There is no maximum length of a message. Clients should show at least 140 characters of a tweet after collapsing the mentions.” Or something like that.
@kas@enotty.dk The amount of whitespace around the equal sign shouldn’t matter. Wouldn’t be a comment above the line not enough? <# nick = mdom # my nick> looks weird.
Horay! This is the most active 24h for #twtxt in the last month. Already more than 50 posts.
@kas@enotty.dk, @freemor@freemor.homelinux.net (re: finger) That’s another neat thing about twtxt, it totally independent of any transport layer. ipfs, zeronet, finger, as long as the protocol has an url we could follow the ressource.
@freemor@freemor.homelinux.net We tackle that problem when it arises. I’m currently following 60 (mostly dormant) users with 500k for all files with an update taking less than 2s.
@freemor@freemor.homelinux.net Or we resort the timline to have the newest entries on top, then you could just request the first x bytes of every feed. But archiving is definitly the pragmatic solution. Get back to me if you want to have that in txtnish. Or maybe someone want to submit a PR?
@freemor@freemor.homelinux.net Yeah, but yet a lot of sites are preparing the jump to http/2.0 with its inbelievable complexity. And even http/1.1 is probably a little bit over engineered.
@freemor@freemor.homelinux.net I think we 100% on the same page about self hosting. I thought the article was more about stopping to cooperate with the corporate web. We need to develop standards with our own priorities forefront.
There are so many alternatives like ipfs or scuttlebutt, but i fear that we loose the simplicity of the old protocols. That in my mind is the main attraction of twtxt.
@freemor@freemor.homelinux.net Is it? I mean it paints a pretty dark picture, but in the end Aral appeal to us to fix the whole mess without any corporate involvement.
@kas@enotty.dk Mhh, only 27% of the feeds i subscribed to, declare an explicit charset in the header. Probably easier to fix that in the clients…
@kas@enotty.dk And to make it even worse, most clients interpret the data as win1252. But does any twtxt client autoconvert to utf8 in case another charset is send? I think it probably okay for every client to assume it’s utf8.
@kas@enotty.dk Mhh, i explicitly allow 60 seconds in my rfc3339 regex. Is there already a bug report for python?
@phil@philmcclure.duckdns.org You mean the leap second in evil.txt? It’s expected to break clients … :) You can just skip lines that you can’t parse. Although it’s a valid date according to rfc3339. Maybe file a bug against coreutils?
@phil@philmcclure.duckdns.org RFC3339 is a subset of ISO8601, your timestamps are fine. We’re not using 8601 as it’s crazy complicated to parse: 2009-W01-1 is a valid represenation for Monday 29 December 2008.
I’m not working on bussard, I’m just IRC friends with @technomancy@technomancy.us and he expressed interest in making scheme available there.
@benaiah@benaiah.me Wow, you’re working on bussard? I wanted to try that game for months! Cool
@mdom@domgoergen.com mostly for fun, but once I have it working more it’ll get integrated into https://gitlab.com/technomancy/bussard.
@phil@philmcclure.duckdns.org Can i add you the we-are-twtxt? And how does the telegram bit works? Never used telegram before … can you chat it up and it will post to twtxt?
@kas@enotty.dk, you were both winner and looser in the last bullseye! :)
@freemor@freemor.homelinux.net I’ve thought you lived in Brazil, but right, Canada has also regions with -04:00 …
@kas@enotty.dk, @skingrapher@skingrapher.com 8ball only checks users that are on we-are-twtxt and it does only checks for tweets since its last run. So it won’t catch new users that had questions for 8ball in the past.
Hi @skingrapher@skingrapher.com, i’ve seen that you forked we-are-twtxt and waited for your pr! :)
@benaiah@benaiah.me Do you have to transform old scheme code to lua or is it just for the lulz?
Weird, nobody has written a twtxt.el for #emacs yet?!
@tx@0x1A4.1337.cx I like the visualization of the timeline. A static timeline of we-are-twtxt would be cool!
Meet the firehose at https://domgoergen.com/twtxt/firehose.txt. It publishes the complete timeline of we-are-twtxt every ten minutes. I’ll use it for my bots, so i don’t have to hit your feeds so often. Feel free to use it too!
@kas@enotty.dk Is that too cold or too warm? Also 9°C in Berlin, feels warm for me… :)
@durcheinandr@durcheinandr.de Just as an example how evil you can get. It’s not the new standard theme. Although…
@kas@enotty.dk Fermented slimy soybeans sounds like a hard sell. Maybe something you have to taste? :)
@allgebrah@www.synkretie.net I like the photo story!
@kas@enotty.dk Heh, cool video but it sounds like a bug? https://www.reddit.com/r/videos/comments/5yeefj/alexa_are_you_connected_to_the_cia/depeufn/
@freemor@freemor.homelinux.net Do you mind if i steal some code from you with attribution?
@tx@0x1A4.1337.cx Is you page publicly accessable? Maybe i should add exporting to html to #txtnish, seems like a thing many users wan to to?
@dave@davebucklin.com You can always come to #twtxt on irc.freenode.org for help. Otherwise sh -x is always helpful.
@freemor@freemor.homelinux.net Actually i forgot to push. Should be online now.
@kas@enotty.dk bullseye just calculates sum(|140-length| for all tweets) / number of tweets
@freemor@freemor.homelinux.net Okay, added the new subcommand.
@freemor@freemor.homelinux.net At least you aren’t under the last three.
@freemor@freemor.homelinux.net You have to set sync_followings to a remote plain text followings file. You can use we-are-twtxt, but any file will do. Better back up your followings before you test that. Seemed to work here, but who knows.
@freemor@freemor.homelinux.net What do you mean by time based? I added syncing with we-are-twtxt in 6c2038 but it simply compares the result against the last run?
@8ball@domgoergen.com will answer any question you ask him if the tweet ends with a question mark and only mentions the bot itself. It checks every 10 minutes.