Hi Karen -

thank you for the feedback. My responses are in line.
ginnie


On 02/15/10 17:15, Karen Tung wrote:
> Hi Ginnie,
>
> Here are my comments:
>
> - Section 3:
> * What does "enable components of the installer to log data without 
> the overhead of output" mean?
What I meant by this was the logging framework should hide the 
complexity of logging to different formats or logging different levels 
of output to different locations from the components.
>
> - Section 4:
> * I think  "log levels" as a type of "event filter".  How is it different?
These aren't really event filters. The logging handlers will be the 
event filters.
I'll be sure to clarify this in the revision.
>
> - Section 5:
> * General comment about this section.  All the requirements are listed 
> here, but there's
> no further explanation on the requirements, and there's also no other 
> section in the
> document that explains why the specified interfaces provided are 
> sufficient to satisfy
> all the requirements.
I'll address this and be more specific.
>
> * 2nd bullet: Remote logging is enabled by default".  Not sure whether 
> this is possible.
> To log remotely, I imagine that we need to have network access, and 
> also need
> to set it up somehow, can it really be enabled by default?  What will 
> the behavior
> be if there's no network connection?
I'm removing this as a requirement.
>
> * 3rd bullet: Since using Python logging is not decided yet, I think 
> it is sufficient for
> this bullet point to say that the logging framework needs to provide both
> a Python and C interface.  The way it is presented now implies that 
> Python will be used.
Thanks that's a good point. At the time we were not sure what the 
interfaces would be.
They will be python interfaces. I'll clarify that.
>
> * 4th bullet: The 3 sentences here sounds to be like 3 different 
> requirements.  Why not
> let each of them have it's own bullet?
Ok.
>
> * 5th bullet: For remote logging, does this mean that the logger will 
> write to the target
> directory of the machine with the specified IP address?  I think we 
> need to capture what kind of setup the remote machine need to do in 
> order for this to work somewhere in this doc.
Yes. I agree. I'm working on a server side component for the revision.
>
> * I am not sure what does "The logging framework allows the user to 
> modify logging level from the command line"
> mean.
The user would use a CLI to locate the components of an application and 
to modify the logging level of a particular component
from the command line. I'm not sure, given the conversation we had that 
this will be managed by the logger. Most likely
not. I'll address this point when I revise the doc.

>
> * It is not discussed in this section, or anywhere else in the 
> document.  Will the logging
> framework allow other types of logging handlers be registered, or only 
> the implemented ones
> can be used?
Any logging handler requested by the application/engine can be used.
>
> -Section 6:
> * For all the interfaces, it would be useful to indicate what is the 
> expected type for the input,
> such as "string/int/object"...etc.. Also, which input is required, and 
> which is optional.
> If it is optional, what's the default value?
Ok. I'll be sure to be more explicit.
>
> * section 6.1, can 2 loggers be instantiated at the same time by a 
> single application?
No, only one logger is instantiated per application. There can be 
multiple handlers.
>
> * section 6.2, What's the purpose of this function?
Find_Component is would be used to locate the components of an 
application to modify the logging level of a particular component
from the command line. Given the conversation we had, I'll probably be 
modifying this.
>
> * section 6.3: I am also not clear how this function would be used.  
> It would be good to
> provide more details.
Ok. I'll flesh this out more.
>
> * section 6.4: Looks like there's no optional/required argument to the 
> actual logging
> functions to indicate which component is generating the messages.  Do 
> callers have
> to manually add that to the message before calling this function?
This would be handled by the Register_Component function (6.3).
>
>
> - Section 9:
> * I am not clear what use case 2 is for.  Is it for a user to easily 
> retrieve some specific
> content of the log file?  Or is it that the user only wants certain 
> type of output,
> so, he/she specify that while calling the application.  The 
> application in turn calls
> the logging framework with those filters.
I need to modify this based on a misunderstanding I had. Thanks for 
pointing this out.
>
> * Section 10:
> - This picture shows that the application is the one that instantiates 
> the logger.
> My understanding based on Sarah's caiman design document is that the 
> logger
> is instantiated by the execution engine.  Then, the application can 
> request to
> get a "handle" on the logger.  The same handle will be passed onto 
> each of the checkpoints
> registered in the engine.  That way, we ensure the application and all 
> the checkpoints
> that are registered to run by the application all use the same logger.
Thanks. I'll correct this.
>
> - What's the "XML Manifest" for?
The diagram was meant to be a more generic representation of the more 
specific examples in the architecture document, so
I included it for completeness.
>
> Thanks,
>
> --Karen
>
> On 02/09/10 10:21, Virginia Wray wrote:
>> Sorry about that. It completely escaped my attention that external 
>> folks would not see it.
>> Now also posted here: 
>> http://hub.opensolaris.org/bin/view/Project+caiman/Logging
>>
>> Thank you for bringing this to my attention.
>> -- 
>>                              
>>              Ginnie 
>>      
>>   
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
<http://mail.opensolaris.org/pipermail/caiman-discuss/attachments/20100224/b28b4638/attachment.html>

Reply via email to