[V3dot10] [ANN] Potential Tools Support for Process for 3.10 and 3.9.1

stephane ducasse stephane.ducasse at free.fr
Sat Jan 6 16:18:03 UTC 2007

>> I do not have the time today to look at what you did but this is  
>> going in the right direction
>> Now of course they was a process in 3.9, I hate when people can  
>> say or even think about that.
>> Have a look at the scriptloader class in 3.9
>> The principle was the following.
>>     we created a list of packages + some possible initialization  
>> actions and cleans that were available on the squeafoundation  
>> squeak  source
>>     then we issued a cs that was published on the server that  
>> would load the script.
>>     for the changes that we could do with MC we issued a simple cs  
>> on the server.
>> So if the changes or MC would not have bug we could reproduce  
>> totally an image strating with an old. we were not able because of  
>> squeak mess (a category is not a package so sometimes you have  
>> circul;ar dependency and also because when you fill in the code  
>> that modify the UI and you do not have atomic loading it crashes).
>> So with atomic loading and MC2 we should be able to get there.
>> Stef
> Thank you so much for filling in some details for me, I know it was  
> silly for me to press on without really considering the earlier  
> process, but it is important to integrate approaches where it makes  
> sense.

Have a look.
I strongly believe that what is cruelly missing right is a notion of  
component specification so that we can load a package and load all  
the required
ones. Universes are one solution when the release is done, but when I  
see all the version of seaside and the interaction with the other  
it seems to me that this would be a really important missing pieces  
of the puzzle. The MCCconfiguration are a step in that direction.
But I should be able to ask a package with one I should load to make  
sure it works (with * or >) as in debian and other package systems.


More information about the V3dot10 mailing list