On Dec 16, 2007 2:26 PM, Emmanuel Lecharny <elecharny@gmail.com> wrote:
> This also opens up another topic: what semanitics do we intend for
> adding the values of one attribute to another attribute which
> potentially can be of another AT? I am leaning towards only allowing
> the addition of vales of an attribute foo to attribute bar if and only
> if foo attribute is an instanceOf() the attributeType of bar. Hence
> why I added this method in just recently.

My personnal opinion is that as soon as the bar AT accept the value
taken from AT foo (ie, the value is valid accordingly to the bar SC),
then I don't see why we shouldn't allow this move. At least, it will
be slightly less restrictive.

OK this is making sense why you asked about whether or not you should check the syntax of the value on addition.  You're suggesting that if there is no problem with the syntax upon adding the value of one attribute to another attribute then it should be OK. 

I'm a bit concerned about doing this because this requires throwing away the normalized value if present, then normalizing the value again with the new attributeType of the target attribute the value is being added to.  Syntax checks cost more than simple attribute type checks in general.  Also most of the time transfers of a value from one attribute to another will occur over the same attributeType. 

I'm beginning to start to favor keeping the attributeType reference in the value again.  I think we need to give it more time before we make the final call on the direction we should take here if we cannot agree.    
 
In any case, I had this pb in mind and this was why I asked if we
should check the syntax when adding a value ... For the
ServerAttribute, I think that it's the developper responsability to
assume this check.

I should have read further.  I'm really now favoring keeping this attributeType reference in the value rather than going with this approach because of the consequences.  I think it will pay off in the end for us. 

Alex