Niall Power wrote:
> On Thu, 2007-03-29 at 11:32 -0400, Dave Miner wrote:
>> Niall Power wrote:
>>>>>> om_getDiskPartitionInfo: why not just take a DiskInfo as the first
>>>>>> argument? And what is meant by "A failure will be fatal in this case"?
>>>>> The DiskInfo structure returned from om_getDiskInfo is a linked list of
>>>>> DiskInfo. Why the whole structure needs to be passed back when diskName
>>>>> is enough to get the partitions.
>>>>>
>>>> If this were the long-term design I'd be pushing strongly to have the
>>>> callers not directly referencing structure members but instead using
>>>> accessor functions, because that gives us much more flexibility in
>>>> evolving the design, so generally I'm suggesting that callers not have
>>>> to grovel around in the structures to use results from one call to make
>>>> another call.
>>> +1 on the accessor functions suggestion. I'm even considering writing my
>>> own wrapper functions in the GUI rather than randomly poking at data
>>> structures.
>> That would be a bad answer from my point of view.  If you're finding 
>> need for them, others will too, so let's put them in the right layer - 
>> doesn't mean you can't write them, though that's something you and Sarah 
>> can work out.
> 
> Arguably yes, that would be a bad answer. What I am aiming for is to 
> minimise the amount of code rewriting in the GUI as orchestrator
> interfaces evolve. The best way for me to do that right now is to write
> some wrappers, even very light ones, so I only have to make the changes
> in one place - and only have to break the gnome coding style in one
> place :)
> If it's presumed that the only realistic consumer of the orchestrator's
> target discovery interfaces is an installer UI type of application then
> I think the interfaces should be moved into the orchestrator layer also.
> I'll post the set of wrappers I've written soon - right now I'm at the
> beginning of trying to populate the partitioning UI using dummy data I
> created based on the current design. As I progress I should get a better
> understanding of how the API can be better refined to suit the GUI.

I am not sure that it is presumed that the only consumer of the 
orchestrator's target discovery interfaces is install UI type of 
application. However, that said, Sundar and I are working on a new set 
of interfaces that do include some of the 'getters' that you are looking 
for. We should have a new document out soon.

Send us your wrapper functions however and we can take a look.

Thanks for all your input.

sarah

Reply via email to