@buttface_9000@PetrKraus42@jorgemanru ah, no worries. sorry if I came off a bit preachy, I'm rather passionate about this stuff. easy to forget it's a human on the other end as well.
@buttface_9000@PetrKraus42@jorgemanru ..that whatever it is can:
- not do any harm to my system, no matter how malformed it is.
- can perform its intended purpose in a performant manner.
both of those benefit from simplicity above all else.
@buttface_9000@PetrKraus42@jorgemanru You are a client connecting and receiving data from arbitrary servers...
was literally the end of your previous reply.
@buttface_9000@PetrKraus42@jorgemanru I had half an essay typed up following up on my other reply on what is bad from first principles about each part of the stack, but thought halfway through that the complaints were so obvious as to be wrote. I guess I was wrong.
@buttface_9000@PetrKraus42@jorgemanru I'm responding to your reply that said if one rewrote the web stack, and approached it from first principles similar decisions would be made. Not sure what you think isn't relevant.
@buttface_9000@PetrKraus42@jorgemanru Voila, thing is basically impossible to misuse, the code that reads it is basically impossible to fuck up. Good luck finding buffer overflow exploits when we don't have a dynaimc index in sight.
@buttface_9000@PetrKraus42@jorgemanru Ideally? fixed size packets so desync/mistaking headers for user data is impossible. Need more data? mark some segments continuations, first specifies how many. one malformed? we skip all of them. Stream size different from what you said? connection terminated, bozo.
@buttface_9000@PetrKraus42@jorgemanru Let's start simple. Whyever would you clean-sheet design a format for requests, or anything really in the entire stack, that you know will be subject to attack and misuse, intentional or otherwise, as a TEXT format you have to parse, with all the error conditions that implies?
@buttface_9000@PetrKraus42@jorgemanru The solutions we have were suboptimal even for their time, but were at least workable for their original purposes. If you'd arrive at the same or similar designs even with the advantage of hindsight I really only hope I never have to work with you.
@DoctorDuck77@thewackfack1 Altair was supposed to do the breaking manuever to LLO for the whole stack, Orion+ESM included. Which is part of why the service module is so wimpy. No idea why, but the consequences are still felt today in the HLSes being more complicated than they'd otherwise have to be.
@dumpaccoun16098@rah_66_comanche Don't forget they had a hell of a time developing BE4, "where are my engines jeff" was a meme for a reason. I don't think they had the organizational experience to have done much better even with significantly more resources
@dumpaccoun16098@rah_66_comanche Yeah, they went with a safer design, didn't push it as hard, and still had trouble with it. But tbf Raptor's kind of a high bar.
SpaceX had the best prop engineer(s) of the prev generation, reputation to attract the next, and leadership to take advantage of that. White swan.
@dumpaccoun16098@DudeNumero@antosh_1 Ah yeah, been a while since the last recovery attempt, you're right. I think that's perf related though rather than just them giving up, but I may be mistaken. In light of Starship looming on the horizon and FH not launching that often that seems reasonable as well.
@DudeNumero@antosh_1 Not saying it's unreliable or it can't be done, but there's a lot that can go sideways in unexpected ways. Probably not safe to assume F9 reliability translates 1:1, but there is likely a strong correlation.
@DudeNumero@antosh_1 Center core needs to be different enough structurally, and they haven't had the best luck with recovering those either. Less flight experience since F9 ate its niche (reusabe FH overlaps significantly with upper end F9, expendable is too expensive)