Notices by made out of flesh or wood (nightpool@cybre.space), page 5

@gargron probably not

@lunch pong!

@astraluma @iliana depends on what you're trying to do. if you're trying to keep automated traffic out of your site, recaptcha isn't helping you anyway.

@prophet_goddess "wig"

@Gargron doesn't work anymore :(

@grainloom @chr done (and the other instance @mirzaba found)

@grainloom @chr yeah i have them in the reports queue. i'll try to clear that out today.

@faho @fribbledom again, I don't know Andrew and I don't know his workflow or how he manages his time. When I'm replying to email in the morning, it's certainly a smaller marginal cost to toss off a few replies then it is to pay the cost for context switching into code review mode.

and by the same token, I'm not going to walk into a project and tell the maintainer that their workflow is wrong based on "best practices". I don't know their life.

@faho @fribbledom sure, I'm just saying that the overall point does not strike me as particularly strange, depending on the maintainer and the codebase and what they're comfortable with/used to

@fribbledom it's all about striking a good balance. As a maintainer with not a lot of time, 15 1-line PRs have a lot more procedural and cognitive overhead then 1 15-line PR. it's about balancing the fixed cost of reviewing any PR with the non-linear per-line cost of reviewing PRs. there's a cross over point where the factors trade off which one dominates.

@aschmitz @er1n gossip protocols are cool but it feels like they're going to be much more noticeable, network analysis wise. most machines don't do a lot of p2p (and how do you bootstrap a new node?)

SQID CULTURE SQID CULTURE SQID CULTURE SQID CULTURE

@onethousandtwentyfour I think the only place we currently have direct attribute access left is activity type dispatch but I may be wrong (although we replaced a lot of things with first_value_of which is certainly not the correct behavior in a lot of cases)

@onethousandtwentyfour yeah we definitely fixed this for accounts last year or two years ago. our general approach has been "write our code in the most simple and obvious way possible, while being up front that we're a plain JSON processor as defined by the as2 spec, until someone asks us to refactor for interoperability"

there are a number of benefits to this approach

@onethousandtwentyfour ah, no, we handle this is a couple of different hacky ways throughout the code but we don't seem to handle arrays of types correctly currently. pretty easy to change if someone wants it though.

(although it would be harder for us to support something like what I suggested in the socialcg IRC channel this morning, "type": ["Note", "Announce"], but that's also something we could refactor if necessary