[V3dot10] making and recording changes

[stephane ducasse]

>
> I know this will make Stef and Marcus furious again,

No we are also sad that we could not tame that beast.

> but this is pretty much how I felt about 3.9. I didn't have much  
> time to actually do something myself, about the only thing I could  
> contribute was my experience with MCConfigs - big kudos to Marcus  
> and Stef for pulling this through against all odds. The end result  
> (3.9) *is* much nicer than 3.8.

I would really like to use MCConfigs and work on MC2 because this  
would avoid the problem.
The main problem is that sometimes we could not understand why the  
code we load would break.
For example the story with _ and :=. Marcus spent a lot of time and  
at the end we could not reload it because
we needed atomic load. Changing the system while it is running is  
sometimes difficult.


> Nevertheless, having an accountable history of what happened is  
> invaluable. For example, even if we come up with a new sources/ 
> changes scheme it's next to impossible to redo 3.9 starting from  
> 3.8, to get a full history, right?

We have all the code and all the MC so this is not like we do not  
have the data. But this is clear that this
is difficult. Also the problems is the size of morphic like packages  
and their dependencies.

So I think that it would be good to work on fixing the infrastructure.