Steven email: sjh@svana.org
web: https://svana.org/sjh Other online diaries:
Aaron Broughton, Links:
Linux Weekly News, Canberra Weather: forecast, radar.
Categories:
|
Wed, 29 Sep 2004
More on the Linux v Sun discussion - 11:25
Miguel commented about Greg rejecting the Sun guys API stability arguments. I don't know that he rejected them so much as pointed out that that the API is stable in the kernel <-> userspace interface and has been for many years. Kind of like GTK or Mono or something having published API's and having internal structures. There is not much software if any that needs to use internal structures and such with those libraries. In the kernel though if someone has out of tree kernel code it has to keep up with the kernel internal structures. Andrew Morton has talked about this issue at OLS this year as have various other people, code that gets into the kernel will be maintained. Of course the trick then is getting your code into the kernel, to do this you really need to grok Linux kernel culture and work with it. Mikal pointed out there seem to be exceptions where Linus or others appear arbitrary. Such as FUSE which Mikal suggests wont get into the kernel as Linus thinks it is too close to a Microkernel model. Personally I would hope there are good technical reasons FUSE has not been accepted rather than simply saying all file systems should be implemented entirely in kernel space (after all do we really want GMailFS in kernel space?) Of course Linus is only human (unlike Alan (more Linus quotes)) and has been known to allow code into the kernel in a strange manner in the past. Such as when Dave Miller got the bottom halves stuff in a few years ago. (anyone got a link to something about this I wonder?) |