[aida] Scribo and persistency

Michael Rueger m.rueger at acm.org
Sun Feb 10 19:56:21 CET 2008

Nicolas Petton wrote:
> Le dimanche 10 février 2008 à 10:34 -0800, Michael Rueger a écrit :
>> Nicolas Petton wrote:
>>> Repository>>save
>>>         self persistency saveRepository
>>> Repository>>restore
>>>         self persistency restoreRepository
>> Keep in mind that simply saving the complete repository on every change 
>> (or even every x minutes)
> Yes, but this is only an example. The idea is to have several subclasses
> of ScriboPersistency class. This way you'll have the choice. And we can
> save only what's changed after every change.
>> - doesn't scale
>> - isn't safe with concurrency
> So, what do you suggest ?

Good question, I'm not really an expert on the different persistency 
frameworks, just been bitten by the way SqueakSource ignores the whole 
Disclosure: I think image persistency is the dumbest idea since I don't 
know when... Saving your execution environment together with the data is 
just a recipe for disaster, especially when new versions of the 
application come out. How do you migrate your data forward?

The problem is that with the simple persistency like image persistency, 
you just every now and then, or every change, invoke the save, which 
simply saves everything. This is also true to some extent with Magma, as 
far as I know.

If you use finer grained persistency you might need to keep track of 
which objects you changed in a transaction and save them one by one. So 
just using different subclasses might not be enough, your whole 
application code might need to be built so it notes/signals/tracks changes.


More information about the Aida mailing list