Notices by Erin 💽 (erincandescent@queer.af), page 5

@mia my suspicion is that some of the answer is decrabbification

(By which I mean: removing the Rust code that was only used for doing database migrations)

@mia MK13 is probably better, but I trust the ice shrimp developers to continue turning caffeine into faster database queries (esp. if people keep providing them EXPLAIN ANALYZE traces of performance issues)

@lanodan I mean hes a pc98 fanboy

In the datacentre, nobody can hear you scream

(Because of all the fan noise)

@lanodan @joshbressers it's in the zlib repo but not zlib the library

@evan @hrefna @osma @helge I'm pretty sure Mastodon does actually properly maintain the replies collection, FWIW

@evan @hrefna @osma @helge You seem to be discussing something that is (in part) covered by FEP-5624, which is being led by the Mastodon team

@lanodan it's opaque, generally a struct, implementation varies heavily

The most cursed implementation I know of is that glibc is saddled with an implementation which is a vtable for a C++ class according to the GCC 2.x ABI because the early C++ stdlib and glibc got a bit too close together

Can you imagine spending €769 on this thing only for it to turn up and have a typo on it?

Apparently we're in a world now where people do brand collaborations on motherboards? Ok, wild, but I guess not too surprising.

But managing to typo the brand you're collaborating with is deeply embarrassing...

@chjara 64GB is excessive, but when the machine is 5 years old I expect us to have reached the point where 16GB is very cramped

@chjara I concocted a situation where I was getting a ~50% discount at Apple so I decided to go all in on an M1 Ultra which will be my primary laptop for a long time

@mybarkingdogs @mia if everyone followed the rules you proposed, the only adults that children would have to talk to would be their parents and groomers

And every single victim of grooming that I know has awful parents.

@chris we run glitch-soc

@gsuberland @ariadne @algernon also why has he decided that the age-old base-128 variable-length quantity (that dates back to at least ASN.1 BER, probably longer) format requires rebranding as Kim?

@ariadne @algernon

The representation of numbers is completely independent of obsolete number formats like IEEE 754.

:blobfoxdisgust:

The representation of counts comes from Kim. Kim is also the representation of characters, being simpler and more compact than UTF-8.

:blobfoxcofeglare:

Nota floating point represents numbers as coefficient * power(10, exponent)

:blobfoxdeadinside:

@lanodan its not perfect but it works quite well, and for lots of systems the statelessness is a feature (you should do HTTP/2 with connection pooling though)

But one problem, IMO, is that you end up mixing up object identity and method in your path

@lanodan oh im not thinking of anything complex here

Just when receiving a POST into the inbox with a HTTP Signature which isn't same-origin with the activity being posted, rather than discarding the activity one should re-fetch it from origin because it might be outbox forwarded

(of course this only works if the activity has a dereferencable URI)

@lanodan I think SOAP's kinda cute in some regards but XML is totally the wrong carrier for RPC

(WSDL is actually a good idea, very poorly implemented. And the SOAPAction header is actually, IMO, sort of the missing piece of the RPC-over-HTTP puzzle)

@lanodan I kinda see a use for signing activities so when I reply to this post, your server can relay the creation of my reply (of course the contents of the post needn't be signed, so you refetch that)

but tbh it would be much better if servers just took receipt of a bare URL (or activity with ID out of scope of the HTTP sig) in the inbox as an indicator that they should go and fetch from origin