From: Danny Wyatt (danny@cs.washington.edu)
Date: Mon Apr 26 2004 - 11:32:52 PDT
One of my "resume line" achievements in industry was moving a dot-com
from ObjectStore to Oracle. (This was in 1999.) The entire development
team felt that devising our own object-relational mappings (all of our
software was written in Java) was preferable to dealing with the
complexities ObjectStore introduced to our code. We actually found the
"impedance mismatch" between data persistence and application logic a
nice boundary to have in our design: we wrote a CORBA service for
persistence that handled the mappings and queries and we were free to
remove ODI's interfaces and inheritance hierarchy completely from our
code. So rather than finding an OODB convenient, we found it too
intrusive. Additionally, we felt that the set-theoretic background of
RDBMS's was more well-understood than object-persistence. We had
trouble ourselves deciding how best to serialize our objects so we
appreciated doing away with them conceptually when it came to storage.
As others have mentioned, this paper does seem to miss the importance of
the Internet directly, but I think they touch on some topics that
suggest that direction. The most important of these is the "highly
integrated client component". Looking ahead to 2006, if current
predictions (not those of the paper) hold, then most filesystems will be
replaced with RDBMS's and CORBA-like web services will be the norm for
software architectures. In such a world, the distinction between client
and server is an arbitrary defined only by canonical ownership of the
data. Pieces of data must be continually ferried about and
synchronized, and queries must be distributed to many sources. It is
due (perhaps) to that that XML has risen to such popularity
since--independent of it's support for something like ADT's--it is a
wire protocol useful for data interchange.
This archive was generated by hypermail 2.1.6 : Mon Apr 26 2004 - 11:32:58 PDT