home *** CD-ROM | disk | FTP | other *** search
- Path: news.byu.edu!news
- From: "Stephen W. Liddle" <liddle@byu.edu>
- Newsgroups: comp.object,comp.lang.c++,comp.software-eng
- Subject: Re: Design language?
- Date: Fri, 22 Mar 1996 18:50:55 -0700
- Organization: Brigham Young University
- Message-ID: <315358FF.C79@byu.edu>
- References: <31437B51.27C3@bitmailer.net> <4i6qhq$4e8@gaia.ns.utk.edu> <4i9s82$4ps@Starbase.NeoSoft.COM> <314E0DD6.73F1@byu.edu> <4isdl7$ccu@Starbase.NeoSoft.COM>
- NNTP-Posting-Host: swliddle.byu.edu
- Mime-Version: 1.0
- Content-Type: text/plain; charset=us-ascii
- Content-Transfer-Encoding: 7bit
- X-Mailer: Mozilla 2.0GoldB2 (Win95; I)
-
- Tim Dugan wrote:
- >
- > In article <314E0DD6.73F1@byu.edu>, Stephen W. Liddle <liddle@byu.edu> wrote:
- > >Tim Dugan wrote:
- > >>
- > >> In article <4i6qhq$4e8@gaia.ns.utk.edu>,
- > >> Matt Kennel <kennel@msr.epm.ornl.gov> wrote:
- > >>
- > >> >I think the reality is that once some ideas and technology get settled
- > >> >and reliable enough to be automated, they end up--and should end up--in
- > >> >the ordinary language. [...]
- > >>
- > >> [...] over time, we will add enough
- > >> formality and details to systems like the "unified method" to make
- > >> them executable and so they will generate complete code sets
- > >> where necessary.
- > >
- > >This is the direction I'd like to push things -- let the model contain
- > >the program. In fact, I'd go so far as to hope that we could avoid
- > >"generating code" at all from the executable model. (Eventually) we
- > >ought to be able to create efficient systems by compiling executable
- > >models directly, the same way we now create efficient programs in
- > >high-level languages and rely on optimizing compilers for efficiency.
- >
- > Yeah...sure...to some extent, how we go from the graphic design to
- > the working software is irrelavent except that, as we improve on
- > a step-by-step basis, we will generate various levels of intermediate
- > code first...even today, many C/C++ compilers generate intermediate
- > assembler code...and we are somewhat constrained by the current
- > linker conventions...
-
- Agreed. But my point is that we ought to be ignoring the middle code
- layer. If we happen to generate C++/Java/Eiffel/... that's fine. But
- we ought to think of system analysis/design/implementation at the model
- level, and bring up the executable specification to that level. I suppose
- you might care about the code layer if you're working on optimization, but
- like Knuth pointed out many years ago, the first rule of optimization is
- "DON'T!" (It's rarely needed.) Bottom line: 3GL code generation ought
- to be invisible to the same extent assembly-code generation is today.
-
- > >"Unifying Modeling and Programming through an Active, Object-Oriented,
- > >Model-Equivalent Programming Language," S.W. Liddle, D.W. Embley, and
- > >S.N. Woodfield, Lecture Notes in Computer Science, no. 1021, pp. 55-64,
- > >Springer-Verlag, Berlin, 1995.
- >
- > Any chance it's available on-line?
-
- Copyright issues here. :( I can e-mail you an earlier working-paper
- version of it if you'd like (e-mail me if you want a copy). We're
- also putting together an expanded version for a chapter in an upcoming
- book on OO data modeling.
-
- Cheers!
-
- Steve Liddle
- mailto:liddle@byu.edu
-