On Tue, Mar 20, 2007 at 02:03:48PM +0100, Jurzitza, Dieter wrote:
> Though I agree that the naming I coose is probably not optimal as these 
> modules are not "persistent" by themselfes, the persistence will last as long 
> as the ABI compatibility will last. But this in turn gives persistence and 
> therefore I still think to use "persistence" in the name is not the worst 
> choice in this regard.

This is not about the name of the directory but about whether it does work or
not.

> Therefore I do not agree entirely with Roberts arguments. These modules 
> belong to 2.6.18.2 in the same way as they belong to 2.6.18.8 or any other 
> 2.6.X kernel that keeps this part of the ABI constant. From what I have 
> learned they are not initially related to 2.6.18.2.

I don't think that it is important whether you agree with me or not as long as
you propose to replace a system that works with another system that does not
work.

> IMHO it is only bad to refer to names of components that are not used any 
> more in your system (what is true in the very moment you do upgrade now). 
> Spoken from experience I can hardly remember a SUSE distribution that did not 
> upgrade the kernel at some point in time. So I'd say this is something that 
> should be expected.
> 
> In a "generic" sense it would be best to start with soft-linked modules from 
> the very begin of a distribution, what is in tune with my statement.
> 
> And one could remove the reference to the individual kernel entirely:
> Call the directory /lib/modules/persistent from the very begin. Maybe one 
> would like to use persistent-default / persistent-bigsmp.

This does not work because it does not allow installation of multiple kernels
with different ABIs.

Robert

-- 
Robert Schiele
Dipl.-Wirtsch.informatiker      mailto:[EMAIL PROTECTED]

"Quidquid latine dictum sit, altum sonatur."

Attachment: pgpRCOxnuub7h.pgp
Description: PGP signature

Reply via email to