Re: feedback on Easy keys and universal property

On 25 Apr 2008, at 17:05, Alan Wu wrote:

> Alan,
>
> Easy keys seems to be very practical. Oracle is likely to implement  
> this feature. Don't take it as a promise though :)
>
> Regarding "easiest keys", the counterintuitive case is valid.  
> However, is that a common case? ;)  Do we truly expect
> many users define things  like that?
[snip]

I don't think it's such a matter of commonality per se, but what the  
least surprise would be when that case arises.

It's quite easy to imagine this case happening esp. now that we have  
inline user defined datatypes. The possibility of inducing equalities  
by totally unrelated changes means that the user is unlikely to  
*check* when that happens. Such silent changes are very very  
difficult to debug or even notice.

(One example is capturing video data or other event related stuff.  
You use datatypes to constraint the "moment", but are also using keys  
to merge data from different cameras. Perhaps you shouldn't do this,  
but I've seen such scenarios.)

If you don't use finite datatypes then everything is fine and we dumb  
down nicely.

Cheers,
Bijan.

Received on Friday, 25 April 2008 17:14:06 UTC