5 ms·
The modern packager’s security nightmare
- usr1106 6y agoBasically the same story as discussed here https://news.ycombinator.com/item?id=25758863 https://news.ycombinator.com/item?id=25758863 but I don't think I saw any great answers ... and not discussed here https://news.ycombinator.com/item?id=25323103 https://news.ycombinator.com/item?id=25323103 (Personally I found the second article even more important.)
- 0xbadcafebee 6y agoI predict Go and Rust will eventually return to shared libraries, partly due to these reasons, and partly because a single application will eventually take up a gigabyte of space. The performance impacts of this will make systems so slow that they'll have to go back to shared libraries.
- guitarbill 6y agothese practices evolved for a reason, and we'd do well to remember them. for every security update that's made harder through static linking, the flip side is not sharing global "state" and breaking. Python has virtual envs for exactly this; DLL hell was a real problem. having the choice would be good though; maybe random CLI tools can be statically linked and have a slower security update cycle, which would be a bad idea for internet-facing stuff.
- letitbeirie 6y agoWould you rather have the attack surface that opens up from static linking or from undefined behavior consequent to breaking changes in your dependencies?
- flqn 6y agoNo, I'd rather maintainers of packages consider ABI versioning and stability. I should never be in a case where a patch to a dynamic library causes an ABI break. Stable interfaces are crucial for a healthy software ecosystem.
- usr1106 6y agoMany ours later the same article was submitted again and spawned more discussion: https://news.ycombinator.com/item?id=26203853 https://news.ycombinator.com/item?id=26203853 I hage always wondered why HN does allow to submit duplicates. It might makes sense years (or maybe months) later, but why hours?