Hi Chris,

On Thu, Nov 19, 2009 at 4:17 PM, Chris 'Xenon' Hanson
<[email protected]> wrote:
> Robert Osfield wrote:
>> I'll hold back from doing any more review until you've submitted the
>> other work that depends upon it.  I still am of the view that it's
>> probably not the right way to tackle the problem in hand.
>
>  Hmm. The other code is already complete. I'll see if I can get the go-ahead 
> to submit it.
>
>>>  I don't believe the ReadFileCallback is a safe way to integrate with the 
>>> pseudoloader.
>> Could you explain why you don't think it's a safe way to integrate
>> with a pseudoload?
>
>  Well, the ReadFileCallback would have to know which files should and 
> shouldn't qualify
> for the suffix modification (something that can only be inferred by 
> determining if the
> node in question sits below an already-pseudoloader-affected file). Some 
> degree of
> scenegraph state would need to be dedicated to marking which nodes should and 
> shouldn't
> receive the treatment, so that the ReadFileCallback could later determine 
> this. So, it
> makes just as much sense to add a dedicated suffix-modifying state indicator 
> (as I did)
> because I'd otherwise have to use (and possibly stomp on existing user use 
> of) some
> generic flag or state indicator on the PagedLOD nodes.
>
>  Also, ReadFileCallback is already a commonly-used technique, and only 
> permits a single
> ReadFileCallback to be registered at a time. This would make the pseudoloader 
> technique,
> itself a pretty broadly-useful piece of code, mutually exclusive to other 
> valuable
> techniques that employ the ReadFileCallback. It doens't make for a nice, 
> modular, general
> solution that can be immediately employed anywhere without side-effects.

My strong objection your suggested change is that it's a mod to the
core OSG for a very narrow type of usage, that not at all typical
usage.  We can't just go bloating the core OSG because each users find
it's a little more convenient to implement their particular usage
model.  One of the mantra's I follow is "Minimal and Complete", any
additions from the minimal really have to be justified.  If we can
implement the functionality in a less intrusive way then this is the
right way to do it, not to pollute the core OSG with yet more API and
member variables and obscure cases.

Robert.
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org

Reply via email to