[Coldstuff] Waifs draft 2 (was: Re: First version of waif-patch)

[xmath]

>So, we thought that it'd be easier to just make it possibly for the 
>methods table to be NULL, only costing 4 bytes (by also moving 
>idents and strings into a single structure for the method data).

Sounds very nice

unless you think that's too ugly, you could union it (since it's not 
used for waifs) with something that's waif-only, like the cclass or 
the usage count.


>Brad's mostly done implementing this into the current dev-tree. 
>This has the advantage of also reducing the size of any object 
>without any methods fairly significantly. (About 1316 bytes or so, 
>per object.)

Nice, nice



>should significantly reduce the number of bugs

Actually, I found that changing the name of the struct was fairly 
useful: it caused a lot of compile errors that forced me to check 
every single occurrence for the impact of waifs.


>as well as the complexity of the patch that will be needed.

That is true.


>I'm sorry that this comes after your two efforts, but it looked like 
>the cleanest way and I guess that Brad was bored and in the mood to 
>do the implementation work on that part of it.

Don't worry, I sometimes throw away many versions of my own work 
until I'm satisfied with it. I'm just trying to get waifs done in a 
hurry because I need 'em for trying out other ideas. :)


>If this works for you as an implementation basis, let's talk about 
>some of the rest of the changes.  I think that this may make some of 
>them easier with an object and waif using the same data storage.

Yes, definitely. Until the dev release is available, I'll work on other things.

  - xmath