To view parent comment, click here.
To read all comments associated with this story, please click here.
Did he in fact recommend this? Links?
All I've ever heard from him when asked is "I am only interested in Linux and I don't care about other platforms." I never heard a more rational response than that.
if you would read mailing list instead of just posting here, you wouldn't need links. you can read and search your self. but here is a quick review
he stated following:
- systemd proposes interfaces where DE can integrate tightly into system core
- he is prepared to separate interfaces from core in order to ease the portability, but he is not prepared to accept patches for other than linux inside his core branch. interfaces are portable and he is not prepared to work on anything that is not linux.
- every other os should implement their own clean implementation of interfaces to avoid getting implementation riddled with #ifdef. this also means their implementation will be clean and easy to maintain
- all that os needs is to prepare negative fallback and later implement it in correct fashion
this is how i understood his proposal, but i might be wrong. major mistake he made in proposal was subject line. everything else is really great
what bugs me at this topic is that he proposes one thing that would very much save the day when udev/devd fiasco came (instead of udev implementation, only interfaces need to be implemented trough devd), but BSD people didn't like that one... and guess what, they don't like this one
All I've ever heard from him when asked is "I am only interested in Linux and I don't care about other platforms." I never heard a more rational response than that.
You don't need me to supply links - just try the ones in the article. Lennart repeats the point in almost every reply he's made on that thread...





Member since:
2008-08-19
Be fair - systemd is a *very* low level project, just one step removed from the kernel, and very heavily tied to kernel interfaces. Trying to make the *code* portable just isn't feasible, which is why Lennart instead recommends writing API-compatible equivalents for BSD or Solaris.