Re: Assembly language (was: Re: Another 1.3 wish.)

[sdl@linus.UUCP (Steven D. Litvintchouk)]

Posting-Front-End: GNU Emacs 18.47.1 of Sun Aug  2 1987 on linus (berkeley-unix)



In article <2176@xanth.UUCP> kent@xanth.UUCP (Kent Paul Dolan) writes:

> DOD could save grunches of training costs downstream by sponsoring
> high quality, validated, PUBLIC DOMAIN Ada(tm) compilers for the
> existing suitable home computers now.  

The DoD already sponsored the development of two Ada compilers +
associated toolsets: the Army Ada Language System, and the Air Force
Ada Integrated Environment.  For various reasons, these were not
entirely successful, especially compared to the commercial efforts.

> The Mac, the PC-AT and clones,
> the Amiga 500/1000/2000, and the Atari come to mind as likely targets
> for such an effort.  This would probably be a direct 100 or 1000 to 1
> benefit to cost ratio in terms of DOD and other government training
> money saved by having folks train themselves in Ada, and would aid the
> entire national software productivity picture by vastly upgrading the
> use of a maintainable, software engineering oriented language
> nationwide, as a no added cost side benefit.

I agree wholeheartedly.  The Amiga is especially interesting because
of its multitasking exec built into hardware.  With multitasking
supported by message passing, the Amiga might provide a highly
efficient runtime environment for Ada.  Have any Ada compiler writers
looked at the Amiga either as a host or target?

> At the rate things are going, it might be 5 years before a validated
> Ada compiler priced for the home user is available.  That is just five
> more years of DOD funding essentially all Ada training.

Meridian is developing a compiler for the IBM PC, and it's logical to
assume that they or someone else will host/target the Mac.

> The excessive (better, not sensitive to company size) cost of
> validation probably prevents a lot of small companies from considering
> making a splash in the Ada compiler pool.  (The 20 man years or so of
> high priced talent required doesn't help a lot, either, of course.)

Perhaps another reason is that Ada compilers wouldn't compare
favorably with Turbo Pascal, Manx C, etc., because of:

   a.  Efficiency:  I can get a Modula-2 compiler for my Amiga that 
	generates code comparable in efficiency to C.  Also, I can
	run everything, libraries and all, off a single floppy.
	What Ada compilers can do a comparable job in compilation
	and runtime efficiency?  Will I be forced to buy a 50 megabyte
	hard disk to host the Ada libraries on my Amiga?

   b.  Target environment:  Again, the Modula-2 compiler for my Amiga
	provides a *full* language interface to the graphics routines, the
	windowing interface, the Amiga exec, the ROM routines, etc.
	Essentially I can program nearly anything in Modula-2 that
	I can program in C.
	But most Ada compilers don't provide such wide
	interfaces to the target machine.
	It's ridiculous for Alsys, say, to sell a PC compiler
	that takes over the whole machine, bypasses MS-DOS, and
	requires a special board.
	
There are two kinds of Ada users: those who program in Ada because the
DoD tells them to, and those who program in Ada because they genuinely
feel it's a superior programming language.  If you want to reach this
latter group of users, you must provide the same kinds of efficiency
and support facilities that people have come to expect from C, Turbo
Pascal, etc.  The excuse that it's OK that Ada compilers consume a lot
of resources because they're doing so much work to process *huge*
multiperson software wears a little thin in the PC world; I don't
write huge multiperson programs on my Amiga.


Steven Litvintchouk
MITRE Corporation
Burlington Road
Bedford, MA  01730

Fone:  (617)271-7753
ARPA:  sdl@mitre-bedford.arpa
UUCP:  ...{cbosgd,decvax,genrad,ll-xn,philabs,security,utzoo}!linus!sdl