[cap-talk] Styles of persistance

[Jonathan S. Shapiro]

On Thu, 2008-04-03 at 18:51 -0700, Jed Donnelley wrote:
> Hmmm.  I'm not sure I'm understanding the distinctions above,
> e.g. when BillF says "Serialization is not used to move it from
> system to system. :-)".  Objects of course vary in terms of their
> internal complexity.  When the stored version of an object needs
> to change from one version of a server (system?) to another,
> I can't see how one can avoid doing a conversion - read in the
> old version and write out the new.  I took that to be what Jonathan
> was referring to as "serialization".  Of course the object might
> not be literally serialized down to bits (bits and capabilities in
> descriptor based systems?), but they do have to be reformatted I
> think as that seems to be the whole point.

The space bank is an active object that must remain operational during
conversion. Conversion is likely to require storage allocation, and
storage is provided by ... the space bank. So conversion is converting a
moving target. 

In light of this, one would probably NOT serialize the space bank. One
would inject a front-end process and then do a data migration from the
old bank to the new.

shap