3 ms·

Another reason software sucks is that developers are sometimes so incredibly bad at communicating their ideas (which impacts documentation, development planning

Another reason software sucks is that developers are sometimes so incredibly bad at communicating their ideas (which impacts documentation, development planning (especially in the age of distributed open source projects), and even code quality [in terms of naming, commenting, etc]). "I don’t have much involvement in portable, but I definitely had a hand in neutering the RAND and egd interfaces. Contrary to some commentary, we didn’t neuter these interfaces because we didn’t know what they were. We neutered them because we know precisely what they are. They’re fucking stupid." uh... okay; thanks for the enlightenment. I'm sure this post makes more sense to people who are intimately familiar with whatever it is that is being talked about in the first place, but to the rest of us it might as well be written in Klingon.

yellowapple 12y ago
Having read the various slides on why libressl is deemed necessary, it makes quite a bit of sense. A large chunk of OpenSSL's API has no business existing in the first place, let alone in a public scope, because said chunk consists mostly of redundant and half-baked implementations of things that exist in the operating system and/or external libraries. OpenSSL took a rather strange approach of "hey, $some_obscure_operating_system doesn't provide $some_critical_system_function, so let's implement it ourselves and activate it for every single platform that OpenSSL is used on regardless of whether or not there's an existing equivalent". In many cases, these sorts of shenanigans are the source of various bugs or - just as horribly - the reason why so many bugs had been undetectable via stricter memory access requirements and debugging techniques (the latter being what's referred to when the phrase "exploit mitigation countermeasure" is used in the context of OpenSSL). Ted could probably be a bit more helpful by providing a link to the "fucking stupid[ity]" that is OpenSSL's bass-ackwards API, but most of this is already explained in prior announcements about libressl's improvements/development and such, so it's understandable why he didn't feel the need to provide links for context.

yellowapple 12y ago

Having read the various slides on why libressl is deemed necessary, it makes quite a bit of sense. A large chunk of OpenSSL's API has no business existing in the first place, let alone in a public scope, because said chunk consists mostly of redundant and half-baked implementations of things that exist in the operating system and/or external libraries. OpenSSL took a rather strange approach of "hey, $some_obscure_operating_system doesn't provide $some_critical_system_function, so let's implement it ourselves and activate it for every single platform that OpenSSL is used on regardless of whether or not there's an existing equivalent". In many cases, these sorts of shenanigans are the source of various bugs or - just as horribly - the reason why so many bugs had been undetectable via stricter memory access requirements and debugging techniques (the latter being what's referred to when the phrase "exploit mitigation countermeasure" is used in the context of OpenSSL). Ted could probably be a bit more helpful by providing a link to the "fucking stupid[ity]" that is OpenSSL's bass-ackwards API, but most of this is already explained in prior announcements about libressl's improvements/development and such, so it's understandable why he didn't feel the need to provide links for context.