Helium is quite difficult to contain. It's the smallest elemental molecule (yes, smaller than hydrogen, even though it's heavier) so it can slip through the tiniest of pores.
I'm only a layman in this area, but I find it interesting that NASA and Boeing/ULA seem to often have a principled approach to these sorts of things, e.g. using helium, hydrogen, hypergolic fuels, aluminium fuselages, realtime computer systems etc, whereas SpaceX seem to have a more pragmatic approach with methane, steel, etc.
The "traditional" fuels make sense (chemically), may provide technically better stats, or hypergolic fuels may be easier to make work in some way. Aluminium is lighter, so seems like the right choice. However, in practice, methane may be technically less capable, but much easier to work with. Steel is worse than aluminium in many ways, but it's also strong and easy to manufacture with. And where the shuttle was famous for having a complex (but well engineered) sophisticated avionics system, SpaceX use "commodity" computers running Linux.
SpaceX seem to have had a lot of catching up to do with some of this, but it seems to have paid off, or is paying off over time with what is now reduced complexity.
Pressure-fed engines are the simplest liquid-fuel engine design [1]. SpaceX's Raptors' full-flow staged-combustion cycles are massively more complicated. Yes, keeping your helium from boiling off is...a job. But taken as a system, the Raptor is more complicated.
> methane may be technically less capable, but much easier to work with
Also less precedented. Hydrazine/helium thrusters are amply studied. On technological risk, methane is untested.
> Steel is worse than aluminium in many ways, but it's also strong and easy to manufacture
It's also heavier and corrodes.
In summary, SpaceX chose a more-complicated engine design, a novel fuel and an unprecedented structural material and beat Boeing to the punch. In part because Boeing sucks. In part because Starliner has been obvious fuckwit fodder for at least half a decade, and that leaves you with a special sort of person who is willing to work on it.
> where the shuttle was famous for having a complex (but well engineered) sophisticated avionics system, SpaceX use "commodity" computers running Linux
The Space Shuttle's main computer was a System/360 descendant [2]. Its design philosophy implies that they would have used Linux were it an option.
Starship and Starliner aren't really in competition though. the SpaceX Analog is the Falcon 9 and Dragon. which at its fundamental level is a pretty simple spacecraft. The Merlin engines in the Falcon 9 are Kerosene Liquid Oxygen Gas generator engines. which is one of the more simple basic designs you can choose for a rocket engine. The thrusters are pretty standard Hypergolic thrusters.
> Merlin engines in the Falcon 9 are Kerosene Liquid Oxygen Gas generator engines
Still far more complicated than a pressure-fed engine.
The point of the comparison is in the benefits of SpaceX’s economies of scale. They’re building Raptor engines that are failing at a lower frequency than Boeing’s pressure-fed hyperbolic RCS.
These System/360 descendants were programmed in languages I've never heard of, such as JOVIAL and HAL/S. It feels that people were having all the CS fun in the 70s and didn't have to deal with the garbage pile of Javascript, "Modern" C++, flaky frameworks etc. Am I romanticizing?
