On Mon, 21 Feb 2005 10:24:31 -0800, Tim Rowledge <tim at sumeru.stanford.edu>  
wrote:

> Please don't forget that for a long time now 4.0 has been assumed to be
> the point at which we break image backwards compatability and move to a
> new object format (at least for compiled methods) and clean up a lot of
> old back-support VM code.
>
I know. It could well be that what we've got in mind for the  
version-after-3.9 will eventually turn out to be 3.10 or something. The  
idea is, in any case, to do the image cleanup stuff pretty soon, and even  
though no VM changes will probably be introduced in that version, it's  
still attractive to bump to 4.0 - a base image with RestOfMorphic+Etoys  
not there, a freshly written V4 sources file, and an almost empty change  
set will make the download size of 4.0 quite a bit friendlier. I think 4.x  
would then be reasonably short-lived, and 5.0 would introduce the VM  
changes, object format changes, etcetera - delaying this a bit would give  
people who take 'image modularity' on their shoulders a bit of time to see  
whether they need any VM/image changes (a kernel image is I think what's  
needed most for the next round, and should be doable in a reasonably short  
time frame).

But whether a version is called 3.10 or 4.0 or 5.0 or X is, hopefully, not  
the most pressing topic at the moment ;).

What is more interesting: what do you think of the proposed roadmap?

Also, time-frame wise (I haven't followed the VM/object memory discussions  
closely), are all proposed changes for 'next major version breakage'  
clear? Is there a list somewhere? In other words, could this be done  
tomorrow or is it a whole project in itself to hash out what's needed?

(apart from anything required by 'modularity', 'kernel Squeak', or  
whatever you want to call it - I don't think one should need lots of  
VM/object memory changes for that but I'm not the specialist here).