eberard@ajpo.sei.cmu.edu (Edward Berard) (10/12/89)
I have been researching the concept of "reflection" in object-oriented systems, e.g.: - The series of papers in the OOPSLA '89 proceedings: - "Metaclass Compatibility" by Nicolas Graube - "Computational Reflection in Class Bases Object Oriented Languages" by Jacques Ferber - "Reflective Facilities in Smalltalk-80" by Brian Foote and Ralph Johnson - Pattie Maes oft cited article "Concepts and Experiments in Computational Reflection" in the OOPSLA '87 proceedings - Nicolas Graube's article in the ECOOP '88 proceedings: "Reflexive Architecture: From ObjVLisp to CLOS" I am interested in accomplishing the following: 1. Developing a definition for reflection which reflects the _concept_, not a particular _implementation_ of reflection. a. I would like a fairly informal, intuitive definition of the concept, preferably not tied to any particular programming language b. I would also like a pointer to a _generally_ _accepted_ (if such a thing exists) formal (mathematical) definition of reflection. 2. Developing a taxonomy of reflection. Specifically, there are types of reflection which are conceptually different from each other, and one, or more, _conceptual_ models for each of these different types of reflection. Of course, taxonomy implies nomenclature, definition, and classification. 3. Developing a rationale for considering reflection. While I will admit that the problem is intellectually intriguing, I am looking for "practical" reasons for considering reflection, e.g., why would your "typical object-oriented software engineer" be interested in reflection? I was in the process of formulating my own set of answers to the above, when I remembered something someone said to me at OOPSLA '89: "We just had a Birds of a Feather session on reflection, and there still seems to be a great deal of confusion on just what reflection is." What are your thoughts on the matter? -- Ed Berard Phone: (301) 353-9652 FAX: (301) 353-9272
twl@brunix (Ted "Theodore" (W) Leung) (10/13/89)
In article <599@ajpo.sei.cmu.edu> eberard@ajpo.sei.cmu.edu (Edward Berard) writes: >I have been researching the concept of "reflection" in object-oriented >systems, e.g.: > - The series of papers in the OOPSLA '89 proceedings: I haven't had a chance to look these over yet, but I will as soon as my copy of the proceedings arrives... > >I am interested in accomplishing the following: > 2. Developing a taxonomy of reflection. I'd be interested to hear about your classification of reflection types > 3. Developing a rationale for considering reflection. At this point, I think that reflection has some applications to building debugging tools and monitoring tools, in addition to the applications mentions in Maes' paper. You might also look at the paper by Watanabe and Yonezawa in the OOPSLA '88 proceedings. They give some other examples. At this point, I think that one of the most distressing problems is that reflective systems are just very slow, because of the overhead involved in maintaining a causal connection between the meta system and the object system. I think that one of the places where reflection will prove to be most useful is the area of self-adapting software. -------------------------------------------------------------------- Internet/CSnet: twl@cs.brown.edu | Ted Leung BITNET: twl@BROWNCS.BITNET | Box 1910, Brown University UUCP: uunet!brunix!twl | Providence, RI 02912
ttwang@polyslo.CalPoly.EDU (Thomas Wang) (10/13/89)
What are reflections?
-Thomas Wang ("This is a fantastic comedy that Ataru and his wife Lum, an
invader from space, cause excitement involving their neighbors."
- from a badly translated Urusei Yatsura poster)
ttwang@polyslo.calpoly.edurh@mitre.org (Rich Hilliard) (10/14/89)
In article <599@ajpo.sei.cmu.edu> eberard@ajpo.sei.cmu.edu (Edward Berard) writes: > I have been researching the concept of "reflection" in object-oriented > systems... > > I am interested in accomplishing the following: > > 1. Developing a definition for reflection which reflects the > _concept_, not a particular _implementation_ of reflection. You want to take a look at the early papers by Brian Smith, particularly his doctoral thesis at MIT, "Reflection and Semantics in a Procedural Language" (1982). Also check out Richard Weyhrauch's "Prolegomena [I love that word!] to a Theory of Mechanized Formal Reasoning" (Artificial Intelligence 13 (1,2) 1980). Both happen to be working in Lisp but that is not particularly important to what they have to say. > A. I would like a fairly informal, intuitive > definition of the concept, preferably not tied to > any particular programming language Here's my intuitive, first order definition: A "reflective system" contains a model of itself which it may access (inspect) and manipulate to change its subsequent behavior. * They call it reflection because that model is in some sense an accurate depiction of itself (like a mirror, right?). * The fact that the system, by manipulating the model, can modify itself requires a "causal connection" between that model and the system. > b. I would also like a pointer to a _generally_ > _accepted_ (if such a thing exists) formal > (mathematical) definition of reflection. I too, would like to know of work on formal definitions, I posted a query to comp.theory to just this effect, yesterday (see below), but there aren't any "generally accepted" definitions. One of the points Smith and Maes and others make is that reflection is "theory-relative" you can obviously only reflect on, inspect, and manipulate things that you can refer to. [My interest in this is that I want to use reflection as an approach to specification. My hypothesis is that if you adopt some form of reflection as a first principle, you can define other descriptive mechanisms in terms of it in such a way that those definitions can be reasoned about. The goal is a readily extensible, thus understandable, specification language whose constructs or features are all readily "provable". But I can talk about this some other time.] > 2. Developing a taxonomy of reflection. Specifically, there > are types of reflection which are conceptually different > from each other, and one, or more, _conceptual_ models for > each of these different types of reflection. Of course, > taxonomy implies nomenclature, definition, and > classification. At a minimum, any such taxonomy will have to consider: 1. What can you reflect on? [already you see distinctions in the literature like computational (procedural, ...) vs. conceptual (declarative) reflection, etc.] 2. When do you reflect? 3. What modifications will the model allow? When can the modifications take effect? PREVIOUS POSTING to comp.theory: Path: linus!rh From: rh@mitre.org (Rich Hilliard) Newsgroups: comp.theory Subject: Formalization of the notion 'reflection' Message-ID: <RH.89Oct12101331@linus.mitre.org> Date: 12 Oct 89 14:13:31 GMT Sender: rh@linus.UUCP Distribution: comp.theory Organization: Mitre Corporation, Bedford, MA. Lines: 33 The notion of reflection, of a system's ability to inspect a model of itself and possibly modify that model to cause a change to its subsequent behavior, goes back to the work of people like Richard Weyhrauch and Brian Smith. Lately, you see "meta-programming" facilities in object-oriented and logic programming languages (e.g., CLOS). I would be interested to know of any work that attempts to formalize some notion of reflective system, what are the essential properties of a system to be reflective, etc. Rich Hilliard The MITRE Corporation - Mail Stop A156 Bedford, MA 01730 (617) 71-7760 Internet: rh@mbunix.mitre.org UUCP: {decvax, philabs}!linus!rh -- Rich Hilliard The MITRE Corporation - Mail Stop A156 Bedford, MA 01730 (617) 271-7760 Internet: rh@mbunix.mitre.org UUCP: {decvax, philabs}!linus!rh
jeff@aiai.ed.ac.uk (Jeff Dalton) (10/25/89)
In article <RH.89Oct13142047@linus.mitre.org> rh@mitre.org (Rich Hilliard) writes: >> 1. Developing a definition for reflection which reflects the >> _concept_, not a particular _implementation_ of reflection. > >You want to take a look at the early papers by Brian Smith, >particularly his doctoral thesis at MIT, "Reflection and Semantics in >a Procedural Language" (1982). If you're interested in Brian Smith's work and want to see a very clear presentation of how reflection can be accomplished in Lisp, you should look at Alan Bawden. Reification without Evaluation. Proceedings of the 1988 ACM Conference on Lisp and Functional Programming. From the abstract; Constructin gself-referential systems, such as Brian Smith's 3-Lisp language, is actually more straightforward than you think. ... In particular, it is not necessary to redesign quotation, take a stand on the relative merits of evaluation vs. normalization, or treat continuations as meta-level objects. Run, don't walk; do not pass Go; etc. Jeff Dalton, JANET: J.Dalton@uk.ac.ed AI Applications Institute, ARPA: J.Dalton%uk.ac.ed@nsfnet-relay.ac.uk Edinburgh University. UUCP: ...!ukc!ed.ac.uk!J.Dalton