| |||||||||||||
|
| |||||||||||||
If API is in flux, how about not exposing it for now?If API is in flux, how about not exposing it for now?Posted Jan 7, 2009 20:06 UTC (Wed) by proski (subscriber, #104)Parent article: Btrfs aims for the mainline If the btrfs API is in flux, how about not exposing it for now? It's not like the filesystem cannot be used without that API. There are also more flexible ways of communicating with the kernel, such as configfs.
(Log in to post comments)
If API is in flux, how about not exposing it for now? Posted Jan 7, 2009 20:48 UTC (Wed) by masoncl (subscriber, #47138) [Link]
Unfortunately, the ioctls are required to use the filesystem. The part that is most in flux is the system used to mount multi-device filesystems.
This is just one ioctl, and keeping it around shouldn't be a problem for compatibility.
If API is in flux, how about not exposing it for now? Posted Jan 8, 2009 17:11 UTC (Thu) by nas (subscriber, #17) [Link] Why can't those ioctls be marked as experimental? It's very difficult to design the perfect API up front. Also, I don't see any need to rename the filesystem. Just merge the darn thing already and mark it experimental. I've got over 1 TB in my desktop machine (not really a whole lot of space nowdays) and I'm willing to devote some space to testing. People who compile their own kernels should be able to deal with filesystem format changes.
|
|
||||||||||||