home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.object
- Path: sparky!uunet!spool.mu.edu!wupost!csus.edu!netcom.com!objsys
- From: Bob Hathaway <objsys@netcom.com>
- Subject: Re: GCC 2.3.2 implementation of signatures for C++
- Message-ID: <1992Dec24.225317.26503@netcom.com>
- Sender: objsys@netcom.com (Object Systems)
- Organization: Object Systems
- References: <1h8cg6INNccc@ector.cs.purdue.edu> <1992Dec23.151615.27346@fnbc.com> <1hbojuINNt6u@ector.cs.purdue.edu>
- Date: Thu, 24 Dec 1992 22:53:17 GMT
- Lines: 40
-
- In article <1hbojuINNt6u@ector.cs.purdue.edu> gb@cs.purdue.edu (Gerald Baumgartner) writes:
- > [On adding abstract types to g++]
-
- First, I'm glad to see abstract types introduced for widespread use.
-
- >A distributed object capability is not done by just adding signatures.
- >Our signatures are compile time type definitions. The compiler uses
- >them only for type checking purposes. At run time, the only type
- >information that's available is the signature table (similar to a
- >virtual function table). It is used to dispatch to the correct class
- >method.
- >
- >To allow passing objects in a distributed environment, you need a
- >run-time notion of signatures. And you need to perform type checks
- >(i.e., signature conformance checks) at run time when an object is
- >sent from one machine to another.
-
- In the general case, yes. But statically not if you have the (remote)
- declarations on the sender's side and compile them there. This appears to
- me to be the same distributed or not, with some additional issues in the
- non-trivial cases that might come up in object-oriented operating systems
- (distributed systems).
-
- I don't have the papers with me, but I seem to recall Emerald has abstract
- types with static checking *and* distributed object-oriented programming
- all in one; I'll have to check. I also seem to recall that distributed
- programming was one of the driving forces behind Emerald's abstract types,
- which provide the (at least static) loose coupling called for by
- distribution.
-
- >There is research done here at Purdue on building a distributed object
- >system for a heterogenous programming environment, but this research
- >is independent of the signature language construct for C++. In fact
- >it is independent of the programming language.
-
- [With hand over heart]
- From my old Alma Matter! Who's doing the work?
-
- bob
- objsys@netcom.com
-