udev problems

Bruce Dubbs bdubbs at swbell.net
Sat May 22 12:33:50 PDT 2004


Matthew Burgess wrote:

>On Sat, 22 May 2004 13:20:41 -0500
>Bruce Dubbs <bdubbs at swbell.net> wrote:
>
>  
>
>>You asked for opinions and you answer with what I perceive as sarcasm.
>> 
>>Have you made up your mind?
>>    
>>
>
>The only sarcastic comment in that was the last sentence.  The rest was
>deadly serious.  I always thought that an LFS system should be usable,
>and I don't see anywhere in the book that mandates a user "MUST" go on
>to BLFS to make their system usable.  If the goal of the book is to
>create an intentionally broken/unusable system then that needs
>addressing very quickly IMO.
>  
>
I'm not sure what you mean by usable here.  What can you do with LFS 5.1 
(or 6.0 for that matter) except build other applications that you want 
or need?  The book is obviously not intended to be broken.  Its not.  It 
is a basic platform from which to build the packages a user needs or wants.

>The main concern I have about your proposal about dynamic device
>creation being a BLFS topic is consistency.  What is so different about
>those types of devices that mean that LFS shouldn't cater for them -
>after all, it's not like they're uncommon in this day and age. Like I
>said before, I believe all devices need to be treated equally. So, IMO,
>they either all get created in LFS *or* they all get created in BLFS.  I
>personally couldn't care less which book they go in.
>

I agree that not everything has a clean distinction between LFS and 
BLFS.  Why don't we put Reiser/XFS/JFS in LFS?  They are supported by 
linux 2.6.   The reason is that thy are not needed by all and ext2/3 was 
selected because it is the most common way to do things.  Today static 
devices are the common way to do things.  Dynamic devices are not 
uncommon, but they are not mainstream yet.  Obviously we need to handle 
devices in LFS or the system won't boot.  We need to pick a default way 
just like we pick a default filesystem.  For now I'm arguing that the 
method should be static devices and convert the book to dynamic devices 
when it becomes a bit more mainstream.

BTW, I feel the book covers devices far too briefly today.  For many 
they are magic:  "Just run this script..."  Sure we tell the users to 
edit the script, but IMO the explainations in the script should be in 
the book proper. 

  -- Bruce




More information about the lfs-dev mailing list