🚀 go-pugleaf

In article <48hl8o$mfq@news-rocq.inria.fr>, tomasic@monica.inria.fr says...
>
>A partial solution to this problem appears in 
>
>@InProceedings{ss94,
>  author =       "Schwarz, Peter and Shoens, Kurt",
>  title =        "Managing Change in the {Rufus} System",
>  booktitle =    "Proceedings of the IEEE International Conference on Data
>                  Engineering",
>  year =         1994
>}
>
>Essentially a new release is considered a new version (in the database sense)
>of the signatures of methods.  Object created under the old release continue
>to access old versions of the code (unless the new version ``patches'' the
>old version -- this is used only for bug fixes).  New objects access new
>versions.  There are some other differences, since Rufus has an implicit
>type hierarchy, but perhaps the idea can be adopted to Java. 
>Unfortunately, the old release must be kept around until all objects
>reference only the new release.
>-- 



What's meant here by "objects created under the <old, new> version"?  If 
"created under" means compiled under then the problem isn't solved at all,
since the problem is precisely the need to re-compile.
If "created under" is a meant as a run-time thing then the question is
being begged -- the point made in the original post is that an object
"running under" the new release is going to resolve to a method that
has been overridden in that new release.




-- 
------------------------------------------------
David Strauss
E-mail: dstrauss@popserver.panix.com 
              dstrauss@lehman.com 
-------------------------------------------------