home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!stanford.edu!agate!ucbvax!efftoo.boeing.com!crispen
- From: crispen@efftoo.boeing.com (crispen)
- Newsgroups: comp.lang.ada
- Subject: Retracting yet another unintentional flame
- Message-ID: <9208262123.AA22664@efftoo.boeing.com>
- Date: 26 Aug 92 21:23:15 GMT
- Sender: usenet@ucbvax.BERKELEY.EDU
- Organization: The Internet
- Lines: 59
-
- I received an email from Joe Wisniewski (which I wish he'd post to the
- group) indicating that I had unintentionally sprouted another flame:
-
- >And if you're using a flat pure-OOD style where everybody can call everybody >else, you might be hosed, too.
-
- I was absolutely *not* saying that this kind of architecture is "good" or
- "bad". I was attempting to say that designs in which types are
- mostly declared in specs containing the subprograms that operate on
- the types AND in which instantiations of these types are flowing
- back and forth between multiple packages, will probably have a
- greater number of files to recompile than other kinds of designs.
-
- Note that neither side of the AND is a bad architecture, IMHO.
- I have used them both (the former much more than the latter) and have
- not felt guilty afterwards.
-
- I am not even saying that ANDing the two together is a bad idea (though
- if someone else were to say so, I would probably not write anything to
- rebut them). The *only* issue I want to raise is the necessity for
- recompilation.
-
- If someone else wants to say that frequent recompilation of lots of
- files is symptomatic of a deeper problem which should have been solved
- in design, I will stay right out of that fight.
-
- What I *will* assert is that the hierarchical directory structure I
- showed in an earlier post maps pretty well to the software architecture
- in a particular domain I'm interested in (i.e., I actually used this
- directory structure on a program or two).
-
- In fact, it may map to that domain so well that it might not be as useful
- in other domains. It occurs to me that creating a directory structure
- that maps to the software architecture you employ on a program is a
- pretty worthwhile activity.
-
- While I'm apologizing for things, I received a note from TeleSoft
- indicating that compiling on a Sun certainly does NOT imply that
- you're using Verdix. I do not believe that the person from TeleSoft
- will mind if I repeat here his assertion that their compiler allows for
- the local libraries that I mentioned as a "good thing" in my earlier
- posting.
-
- I also understand that some people may be sensitive about the use or
- abuse of "types packages". Please consider this another domain-specific
- issue. I am certainly not asserting that types packages are either
- good things or bad things to have. I think that there are some domains
- in which it is very useful to have access to the types defining the output
- interfaces of some component without having to take the operations on
- those interfaces along with them. I also think it's useful to define
- frequently-used record types (e.g., position or velocity vectors)
- and program-specific substitutes for package Standard types without
- having to specify in the same place all the operations that
- might be performed on objects of those types. In fact, defining all
- the operations that might be performed on objects of type Normalized
- might be beyond my current skills. But that's just me. Live and let
- live.
-
- Bob Crispen
- crispen@foxy.boeing.com
-