Hi Sarah,

On 01/13/10 08:32 PM, Sarah Jelinek wrote:
> Hi Jan,
>
> Thank you for the review. My comments inline...
>
>
>
>> Hi Sarah,
>>
>> Great job ! Please see my comments/questions below.
>> I haven't gone through the other comments, so please
>> feel free to redirect me to the appropriate thread
>> where applicable :-)
>>
>> Thank you,
>> Jan
>>
>>
>
>>
>> Also from that paragraph it seems that those two Classes are
>> discjunctive sets, but at the beginning of Chapter 2 it seems that
>> Core Engine Classes are superset of Common Application Classes.
>> I think it might be confusing and thus might need to be more
>> clarified.
>
> No, the Core Engine classes are not a superset of the Common 
> Application classes. Actually, there are three sets of class types:
> -Core Engine
> -Common Application
> -Application Specific(which are not detailed in this document)
>
> If this is unclear, I will look at rewording this.


I see - thank you for clarifying this.


>
>>
>> 2.2 Data Collection
>> -------------------
>>
>> "Provide the ability to translate global installation data into
>> valid Automated Installer (AI) manifest"
>>
>> Should then this component be also in charge of the reversed
>> transformation (populating global installation data from AI manifest) ?
>>
>
> We do this with the ManifestParser which will then store the data in 
> Data Collection. So, no, I don't think we need to have this called out 
> as part of DataCollection.


ok.

>
>> 3. Common Application Classes
>> -----------------------------
>>
>> Should ICT be also included  ?
>>
>
> That's a good question. I am surprised nobody else has asked. In 
> general, ICT is pretty specific to the media we are using for 
> installation.


I agree some of ICT tasks are media specific, but I think there is a set
of post installation stuff which is to be done regardless of mechanism
used for the installation, e.g.

* setting hostid
* activating Solaris partition
* installing grub bootloader
* setting Sparc eeprom properies (e.g. boot-device)
* transferring log files to the target
...

How is it envisioned to accomplish those kind of tasks
in CUD ?


> And, with system configuration changes(which you are working on!), it 
> isn't likely those would be part of ICT any longer. In the long term 
> we want ICT to be more in sync with system functionality, as opposed 
> to the special things we do today. So, for example, we can use IPS 
> packages to set the locale for a system.
>
>> 3.1 manifest parser
>> -------------------
>>
>> Only syntactic validation is mentioned - are we going to
>> consider semantic validation of the manifest as well ?
>
> Yes, I think we will consider this. It seems to me that semantic 
> validation should be a different class though, which is why I didn't 
> include it in this document at this time.

ok.

>
>>
>>
>> 3.2 Target Discovery
>> --------------------
>>
>> * Partitions, both FAT and GPT
>> ->
>> * Partitions, both FDISK and GPT
>>
>>
>> Just curious, what will be the purpose of discovering XEN DomU 
>> instances ?
>
> In general, for "in use" data. In particular, when a xVM DomU has a 
> real device as its backing store.


I see - understood.

>
>> 3.2.1.1 Physical Device Classes
>> -------------------------------
>>
>> Why do we need set_*() methods for Target Discovery ?
>> It seems those would be used by Target Discovery itself to
>> populate data structures ?
>
> These are not strictly required, as it is internal to Caiman, but I 
> like having setter and getter pairs. I have no real heartburn if we 
> don't use the setters though.


I see. I was just curious if there is a special purpose for them.


>
>
>
>> 3.2.1.2 Logical Device Classes
>> ------------------------------
>>
>> What LogicalVolume represents ?
>>
>
> SVM, VxVM, ZVOL logical volume types.


I got it now - thank you.


Thank you for the clarification !
Jan


Reply via email to