home *** CD-ROM | disk | FTP | other *** search
Text File | 1990-04-27 | 1.9 KB | 42 lines | [TEXT/GEOL] |
- Item forwarded by A33 to A34
-
- Item 8136051 24-April-90 14:43PDT
-
- From: PASCOE1 Pascoe, Geoff
-
- To: MACAPP.TECH$ MacApp Technical
-
- Sub: The Great GC Debate Continues
-
- Jeff,
-
- It sounds like you and your group really have your act together if the addition
- of GC didn't cause changes to your application-level source (am I reading you
- right?). With regard to slipping in new memory management strategies- I can
- see how C++ would allow you to tune different classes to use different
- strategies, but I have a few questions:
-
- 1) Assuming that your entire system uses the same memory management strategy,
- doesn't simple procedure abstraction (designed properly) allow you to "slip
- in" a new strategy when appropriate. That is, under this case I can't see a
- whole lot that C++ buys you. At least in this area.
-
- 2) Assuming that you use different memory management strategies for different
- classes (overloading in C++ does provide benefits here), don't you forsee
- conflicts between classes that have references to each but use different memory
- management strategies? I know (or, used to know), quite a bit about GC
- (generation scavenging in particular) and it seems to me that GC's, in general,
- and generation scavenging, in particular, don't deal well with object
- references outside their domain. At the very least, the algorithms would be
- "significantly" complicated because of special cases.
-
- 3) Specifically, what are the features of C++ over a minimalist Object Pascal
- that let's you do this in C++ but not Object Pascal? I'm looking for the
- distilled essence here. I think I know some of them, but I'd like to here it
- from somebody that's actually done it. This data could be very valuable to
- Derek White (he's now designing the new Object Pascal). Maybe we can make an
- Object Pascal that has lot's of advantages, but without the baggage.
-
- Geoff
-
-