[Dev] Interationalization Schema Update

[Phillip J. Eby]

At 12:06 PM 7/20/2005 -0700, John Anderson wrote:
>Wwe could easily measure what the cost of testing attribute types at 
>runtime is and if it's not too bad doing it in the debug build seems like 
>a good idea.

Since the type checking is done at commit anyway, it presumably takes the 
same amount of time as doing it earlier, as long as you're only changing 
the attribute once between commits on performance-critical execution paths.

In other words, if you change an attribute only once between commits, there 
should be no difference in performance at all.

If you frequently change the same attribute on the same item, then of 
course the check will be repeated.  However, I would guess that the most 
common cases of a (value) attribute changing more than once are not on 
critical paths, in which case there's no reason to restrict the checking to 
a debug build.

So, the thing we would want to find out is, do we change the same value 
attribute of the same item a lot?  And when we do, is it on something that 
makes a user-observable difference?  (For example, we obviously change an 
attribute more than once if a user edits it, but a single extra type check 
per edit isn't going to be a problem there because the total additional 
time will be measured in microseconds.)