Hi Jean,

> Sarah Jelinek wrote:
>> Jean McCormack wrote:
>>>
>>> I have posted the Extended Partition support requirements and scope at:
>>>
>>> http://opensolaris.org/os/project/caiman/Extended_Partition_Support/EP_Reqts_Scope/
>>>  
>>>
>>>
>>> Please review and provide comments by 5:00 pm MT Wednesday, July 13.
>>
>> Hi Jean,
>>
>> A couple of comments:
>>
>> Depending on how close we are with regard to meeting our deliverables 
>> we might consider taking this out of 'in scope' for now:
>>
>>    * Move code that involves slice or partition manipulation that is
>>      currently in liborchestrator into libti where it belongs.
>>
>> I agree it belongs in libti, and in the long run with the unified 
>> engine work it will have to be there, but let's weigh the cost of 
>> doing this now as opposed to leaving it in liborchestrator and 
>> extending it within liborchestrator if it is easier for now.
> I agree. It will be moved probably to phase 2.
>>
>>    * Update fdisk to use libdiskmgt extended partition utilities.
>>
>> This will never be in scope. fdisk was modified to use libfdisk for 
>> extended partition support. fdisk itself shouldn't be modified to use 
>> libdiskmgt.
> Sanjay alluded otherwise. If there is code in fdisk that does the same 
> functionality as code in libdiskmgt shouldn't fdisk be modified to use 
> the libdiskmgt code?
>
I don't believe we should modify fdisk to use libdiskmgt. It could be 
done, but I don't believe that is the correct thing to do. Why? fdisk as 
a utility has little need for the data libdiskmgt provides, except for 
the media and partition specific data. libdiskmgt is intended as a 
helper library for applications that need a more global set of device 
information. libdiskmgt does "discovery only" of devices, and then 
builds relationships with regard to the devices and the device contents, 
such as fdisk partitions, or controllers, etc...It also provides 'in 
use' data to consumers who need to know if a specific component of a 
media is already in use by another application so as to not accidentally 
overwrite this data. It is a pretty heavyweight solution for fdisk's 
need to determine if a partition is an extended partition.

-using libdiskmgt to provide extended partition info for fdisk really 
doesn't add value. fdisk uses libfdisk and switching that for 
libdiskmgt, and modifying fdisk to understand the libdiskmgt descriptor 
and association model for getting extended partition data, when libfdisk 
provides this directly seems very invasive.

-libdiskmgt was originally written as a utility to provide its consumers 
with a central point to gather device information. There was no easy way 
to get all the device information from one place. Specific consumers of 
it were SVM's metassist and the CIM device providers. It is still 
intended for those types of consumers which is why the Caiman installer 
libraries use it. It was modified to provide more specific device 
information with the addition of 'in use' checking. In the case of fdisk 
'in use' doesn't mean the same thing and libdiskmgt doesn't provide 
partition 'in use' data. libdiskmgt provides 'in use' data with regard 
to other applications using the resource, such as ZFS, or UFS, or SVM. 
fdisk manages its own 'in use' statistics based on the partition 
information. I believe this separation is the correct separation for the 
type of 'in use' that fdisk needs compared to the other applications.

Again, this could be done but isn't, IMO, the right focus for our group. 
And, I believe that the intended consumers of libdiskmgt does not 
include the fdisk utility.

thanks,
sarah








Reply via email to