Re: Packaging sysfsutils: static library?
Am 2003-10-12 1:44 -0500 schrieb Adam Majer:
> Me too, but I guess I'll have to see how udev looks. How does it look
> like? Will it retain compatability with the devfs device names?
Theoretically yes. The concept of udev was to throw out naming
policies out of the kernel. The userspace daemon udev is supposed to
translate the devices from /sys to /dev. This translation is fully
configurable, thus it would be possible to have your USB printer 1
with serial no 123 always mapped to /dev/laserjet, or to use an
arbitrary naming scheme.
There will be a translation policy which uses the old scheme
(/dev/hda, defined by the LANANA) but I read on lkml that there is
currently lack of interest to create a policy using the devfs naming
But IMHO this is not very relevant yet, I tried udev yesterday, by now
its mostly vaporware.
> You do not have to have a seperate -dev and -runtime package if the
> library is small, especially if it is not used by many applications.
Okay, but as soon as two or more packages are using it, updates had to
be coordinated such that all packages depend on the same updated
libsysfs and are uploaded simultaneously... This seems a bit messy to
me, but if it is common use I will do it like that.
> > Is it reasonable to provide just a static library? Policy 8.3 allows
> > it in principle, but since I'm not very experienced at this, I would
> > welcome any suggestions and your opinions.
> Sure, but is it always nice to have a dynamic library if more than
> one program is going to be using the library.
Well, but since the lib is in a very early state, SONAMES are likely
to change quite often. Thus the idea to create only a static lib until
it stabilizes. Then packages linked against it did not need to be
Thanks for any comments,
 Linux Assigned Names And Numbers Authority