Re: [xml-dev] Quiz: static transformation vs. JIT generation

[Razvan MIHAIU <mihaiu@mihaiu.name>]

Bryan Rasmussen wrote:

>this is why quizzes suck.
>  
>
>>= = =
>>
>>They have the 
>>latest updates coming to them from various exchanges in XML format and 
>>they need to be able to supply the updates to the client on various 
>>platforms: web + mobile + handheld devices.
>>    
>>
>are these the same xml formats coming from the various exchanges? knowing this
>would significantly affect my response. Do these exhanges update their data at
>the same or varying rates?
>  
>
    I do not think that the input format is the same as the output 
format, otherwise they would not speak about transformations.


>>Which one would be the best approach for storing the rate updates ?
>>
>>A. separate xml documents for each share
>>B. a single document for each share
>>    
>>
>
>uhm, isn't this the same thing? if I have 20 shares and I have a seperate xml
>document for each share then I have a single document for each share. 
>
>  
>
    I think that "separate document" should be read as documents with 
different structure (e.g. different associated schemas).
    A single document means that all the shares will be saved in an XML 
document having the same structure.

>>C. static transformations for generating the output XML documents
>>    
>>
>what do they mean by output XML documents in this context, do they mean output
>xml documents that will be consumed directly by the various media, for example
>an xhtml document is an xml document that is consumed by a particular media,
>web
> or mobile or handheld devices, or do they mean output xml documents that will
>be used to generate the media to be consumed. 
>
    Since XHTML is XML you can consider this to be an option.
  

>IF the latter this supposes that
>the exchanges produce xml formats that are not the same and thus an
>intermediary
>transformation is needed to get an xml format that can be used to generate the
>media, but if that was the case then we might assume that transformation of the
>actual media was using answer D (otherwise you would just go ahead and use C,
>but this would require a larger number of stylesheets [if we go with standard
>practice of one stylesheet per format])
>
>  
>
    I think that you are going too deep for this kind of test.

>>D. just in time transformation of output XML documents
>>    
>>
>It seems to me like they consider the process of updating the central database
>and outputting the displays to the target media to be the same processes. I do
>not think this is reasonable, especially if performance is critical. I would
>suppose that updating the database is done as xml updates are pushed to them
>(the description of the problem implies a 'push' solution).
>There is not enough information to determine if I would want to do jit or
>static
>transforms, however I would tend to want to do jit transformations because in
>my
>experience this allows one to build richer media. Also I could definitely
>conceive of scenarios in which jit transforms could, by virtue of only being
>called when required, would give superior performance to static transforms.
>  
>
    You are probably thinking to some kind of cache. (but the question 
did not imply that you can use such a mechanism)


> 
>  
>
>>= = =
>>
>>My answer: B
>>The correct answer: C
>>
>>    
>>
>
>since I'm not quite sure I understand B what did  you take it to mean?
>
>  
>
    I was thinking that having only one type of document structure as 
input will improve performance.
    However, after reading Bob's answer I agree that "static 
transformations" in this case is a good option.

>>    To my understanding 'static transformations' means to generate HTML 
>>pages from an XML source and to feed those pages to all the clients. 
>>'Just in time transformations' means to transform the XML to HTML for 
>>each client that requests the data.
>>    
>>
>
>agreed however the question specifically mentions generating output XML not
>output Html, which caused me to question several aspects of the proposed
>problem. 
>
    You can read XHTML instead of HTML and that would solve the problem.


>Also I would say that JIT transformation also opens possibility of
>tailoring responses to each client need. Hence my saying that I would probably
>prefer to use JIT, when you have a JIT solution it becomes easier to build more
>services etc. on top of it from my experience. 
>  
>
    The problem is performance and not the flexibility of your solution.

>>    Is it practical to apply 'static transformations' to some data that 
>>needs to be processed "real-time" ? If yes, how often would you apply 
>>the 'static transformation' ? If the data changes every second you will 
>>have to apply the 'static transformations' every second !
>>    
>>

>Was this on an IBM quiz? I looked over their sample questions one time. I was
>annoyed. 
>  
>
    It's a quiz for the IBM 141 exam but it is not from their web site. 
I cannot tell precisely because each time I see a question on the web 
that intrigues me I just send myself an e-mail so that later I can take 
a closer look.




Regards,
Razvan