Andy,

> I'll now -1 and uncommit any patches to the littlendian 
> package until this is fixed or someone shows me how having divergent 
> implementations of common functinality is a good thing.  (good luck)

Well, I think if you looked at my patch on DWord.java, you 
won't be unhappy. I don't add any functionality. I add
a Javadoc comment to hint to the developer that the class
is a 32 bit unsigned int. I only have to do this because of
the messed up intValue() method.

> So if fixing these divergents is a big enough 
> *priority* for you and Rainer, well FIX them.
> "Don't leave broken windows."

I agree whole-heartedly. The first fix is to deprecate
intValue() from DWord.java. It is just wrong. A 32 bit
unsigned int should not return a 32 bit signed int.
Can we fix this broken window, or do we leave it?

I think the Javadoc patch is important. I have a second one
coming up for it. The following appears in the Javadoc for intValue()
in DWord.java.

FIXME: Introduce a superclass for the numeric types and make this a method of 
the superclass!

So, I e-mail the list saying this is a bad idea. Why don't we get 
rid of this comment? Will you -1 a patch that takes out erroneous
information from the Javadocs? Seems silly to -1 an attempt to
correct the documentation. As a POI user, I expected more functionality
to be added to /hpsf/littleendian based on this comment. Why can't
I remove it? I see this as a broken window also. I am not adding any
functionality. However, documentation needs to be correct.

I think we are trying to do the same thing here, unify a correct 
implementation of little endian functionality.

My apprpoach differs in that:

1) I want to see a dangerous method deprecated
2) I want to modify the Javaodocs of the duplicate functionality
   to make the docs correct. Leaving erroneous information in our
   docs compounds the problem, it doesn't help.

- Drew

Quoting "Andrew C. Oliver" <[EMAIL PROTECTED]>:

> >
> >
> >
> >I don't understand what you want here.
> >
> 
> Not to end up with a spaghetti mess where every part of POI has its own 
> utility library with duplications of classes that are used throughout
> POI.
> 
> If it doesn't become a priority I'll start obstructing bugfixes and 
> enhancements to classes that should be rolled into this.  It may look 
> like I'm being a pain on this, but I've been patient.  I said "lets do 
> it when you get to it" and well its been a few months.  Marc mentioned 
> something about it.  I filed a bug recently...it was accepted.
> 
> So bottom line, if I have to do it I will, but I'll obstruct this 
> package from growing in the mean time. Not because I'm trying to be 
> difficult but because I think its important that we don't end up with a 
> big mess.  What makes POI different according to most people is that 
> we're well documented and have nice easy to read sources (most folks 
> agree).  The way we got that way was that members of this community made
> 
> it a priority.  So if fixing these divergents is a big enough *priority*
> 
> for you and Rainer, well FIX them.  "Don't leave broken windows."  -- 
>  Otherwise leave them alone.  I'll be there soon.  Quality is a priority
> 
> to me.  
> 
> Don't think I don't appreciate your work, its just this little tumor is 
> starting to grow.  Its time to do surgery.  Its not your fault, its 
> mine.  I should have raised the issues before the code was committed, 
> then we'd have had this fixed a long time ago.  Well I'm correcting my 
> mistake.  I'll now -1 and uncommit any patches to the littlendian 
> package until this is fixed or someone shows me how having divergent 
> implementations of common functinality is a good thing.  (good luck)
> 
> -Andy
> 
> Andy unit status:
> 
> 
> 1. centipede documentation - priority high - wait state for Ken to fill 
> in blanks
> 2. essay on portable code generation with XML - priority medium - active
> 3. code cleanup in POI - priority  high - you're starting to witness 
> this effort
> 4. cocoon best practices example - high
> 5. formulas  - priority high- status - wait state
> 6. other processes - swaping in and out.
> 
> 
> >
> >- Drew
> >
> >Quoting "Andrew C. Oliver" <[EMAIL PROTECTED]>:
> >
> >>Okay dudes, I'm going to start being a real pain and -1ing commits if 
> >>you "Have time" to fix bugs in the divergent LittleEndian than you
> have 
> >>time to combine the functionality into the org.apache.poi.util package
> 
> >>instead of causing a "lets everyone build their own little endian util
> 
> >>classes" in POI.  I don't want to see POI turn into a microcosim of
> the 
> >>Apache whole where everyone builds their own conneciton pool or
> >>whatever.
> >>
> >>-Andy
> >>
> >
> >
> >___________________
> >Drew Varner
> >[EMAIL PROTECTED]
> >
> 
> 
> 
> 



___________________
Drew Varner
[EMAIL PROTECTED]

Reply via email to