RPM needs to go on a diet.

[Michael Favia]

Jeff Johnson wrote:

> Kyrre Ness Sjobak wrote:
>
>> tor, 17.02.2005 kl. 04.47 skrev Jeff Johnson:
>>  
>>
>>> Kenneth Porter wrote:
>>>
>>>   
>>>
>>>> --On Wednesday, February 16, 2005 1:27 PM +0200 Panu Matilainen 
>>>> <pmatilai at welho.com> wrote:
>>>>
>>>>     
>>>>
>>>>> The kernel package runs hardlink on /lib/modules/*/build/ in %post to
>>>>> save space on FC2 and FC3, that's what's taking so long when 
>>>>> installing
>>>>> and removing kernels.
>>>>>       
>>>>
>>>> What does hardlink do? There appears to be no man page or other 
>>>> documentation for it in kernel-utils.
>>>>     
>>>
>>> Hardlinks's identical files to save space.
>>>   
>>
>>
>> Why are hardlinks chosen over symlinks here?
>>
>
> Look there's a really simple answer to hardlinks and rpm and more:
>
> Choose a kernel, use that.
>
> Chose another kernel, use that, get rid of previous.
>
> With one (or very few) kernels installed, hardlink and rpm performance 
> is a non-issue.

At a cursory glance, arbitrarily limiting the ability to rollback kernel 
upgrades (to a small or larger extent) seems a poor fix. Discovering 
kernel flaws or bugs doenst necessarily happen the first time you use 
the new kernel and may lay hidden for some time. In my company dictating 
acceptable use (outside of normal operating bounds, and without very 
good reason) is seen as a red herring of poor design (no implications 
intended). I suppose you could just go find and download the older 
kernels to test but i honestly dont know where to find archived versions 
of kernel updates (that is: not base and not most recent update). 
Perhaps i am the only one who doesnt know where to find such things or 
maybe information on that subject will reduce the need to keep the 
kernels installed. I dont know the issues involved but i hesitate when i 
hear answers like the above suggestion to problems like these.
-mf