Re: On Legacy Applications and Previous Work

pardo@cs.washington.edu (David Keppel)
Thu, 24 Mar 1994 03:56:16 GMT

          From comp.compilers

Related articles
On Legacy Applications and Previous Work PAUL@TDR.COM (Paul Robinson) (1994-03-06)
Re: On Legacy Applications and Previous Work donawa@bnr.ca (chris (c.d.) donawa) (1994-03-21)
Re: On Legacy Applications and Previous Work bill@amber.csd.harris.com (1994-03-14)
Re: On Legacy Applications and Previous Work baxter@austin.sar.slb.com (1994-03-16)
Re: On Legacy Applications and Previous Work steve@cegelecproj.co.uk (1994-03-22)
Re: On Legacy Applications and Previous Work bart@cs.uoregon.edu (1994-03-23)
Re: On Legacy Applications and Previous Work pardo@cs.washington.edu (1994-03-24)
Re: On Legacy Applications and Previous Work bill@amber.csd.harris.com (1994-03-25)
Re: On Legacy Applications and Previous Work mboucher@silver.sdsmt.edu (1994-03-29)
Re: On Legacy Applications and Previous Work bill@amber.csd.harris.com (1994-04-04)
| List of all articles for this month |

Newsgroups: comp.compilers
From: pardo@cs.washington.edu (David Keppel)
Keywords: tools, bibliography
Organization: Computer Science & Engineering, U. of Washington, Seattle
References: 94-03-034 94-03-060
Date: Thu, 24 Mar 1994 03:56:16 GMT

baxter@austin.sar.slb.com (Ira Baxter) writes:
>[One model of resue...]


For another intriguing view on code reuse see Gregor Kiczales' work on
Open Implementations.


%A Gregor Kiczales
%T Towards a New Model of Abstraction in Software Engineering
%J Proceedings of the International Workshop on Reflection and
Meta-Level Architecture
%D November 1992
%C Tokyo, Japan
%E Akinori Yonezawa
%E Brian Cantwell Smith
%P 1-11


One important lesson here is that implementations typically "lock down"
performance decisions. For example, the implementor uses an array lookup
instead of a <value, pair> list because an array is good for the expected
case dense values. However, clients that have sparse values will pay for
the array implementation with big memory allocations and poor reference
locality. These decisions thus mean that various acceptable *interfaces*
have inadequate performance and that eventually limits the code's
reusability.


;-D on ( Legendary applications ) Pardo
--


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.