Re: Ada9X, Dispatching, and Distribution (Was : Why I like C++)

Re: Ada9X, Dispatching, and Distribution (Was : Why I like C++)

[agate!howland.reston.ans.net!usc!cs.utexas.edu!uwm.edu!ux1.cso.uiuc.edu!u]

In article <C7A1pD.9F3@inmet.camb.inmet.com> stt@spock.camb.inmet.com (Tucker T
aft) writes:
>>I also noted the intro document seems to assume that run-time dispatch
>>always occurs in the presence of tagged types.  
>
>It would be helpful to know what words in the document gave you
>this impression.  In Ada 9X, run-time dispatch only occurs
>...

It was really just a minor point and I see I may be wrong, a passage from
section 2.2 on message dispatch based on class wide types:

"In this case we don't know which procedure to call until runtime because
we do not know which specific type the alert belongs to."

If Process_Alerts is separately compiled, you're right since A's type isn't
known.  I *was* thinking of any of the optimizable PE cases, such as with
value propagation.  Say when:
  red_alert: High_Alert;
  ...
  Process_Alerts(red_alert);

Maybe pointing out that often the dynamic cases will run with the same
efficiency as the static cases (e.g. comparable to C; even ML does this
in a restricted sense) may help to sell the idea to those primarily
concerned run-time efficiency.

>> . . . Chamber's thesis shows
>>...
>Again, suggestions for improved wording are welcome, though we want to
>keep the introduction relatively simple.

I've been working on getting 25 hours out of every day but I've been
settling for 17 lately:-)

>The (Annotated) Ada 9X
>Reference Manual is available for those interested in all of the
>gory details.

Just got it (what a big one).  Does ANSI and ISO require all of that?
No wonder C++ is taking so long.  A co-worker asked when Ada would add
reflection too, since its the next logical step after adding OO.  He
also wondered if it would be desirable to expose all those extensive
innards, saying some wonder if Ada's really just a small language waiting
to get out.  I replied both are true, since a small well-defined reflective
kernel could be used to build Ada!  Ada 2003 maybe, how about this year?
How about a standard...

[Diana comes to mind but I don't know if anyone is using it.]

Are minor and/or major changes to the 9x standard still possible?  How will
the revisions work; will the standard be decided on at the 9x conference this
year?

>>I also would have liked to see something like Argus' non-monitor (multi-
>>threaded) objects added to Ada's active object repertoire; oh well...
>...
>implementation model, and the alternative implementation
>models could in any case be implemented efficiently 
>using the existing primitives of the langauge.

Good point.  It was just one of those experimental features attempting to
improve the tasking model.  I assume you've already covered the Concurrent-C's,
Concurrent Smalltalk's and etc. too.

>In version 2.0 of the Annotated Ada 9X Reference Manual (AARM;2.0)
>on ajpo.sei.cmu.edu (public/ada9x/mrtcomments/rm9x/v2.0/ils.aarm.ps)
>you will find a version of the Distribution Annex (Annex I).
>There are earlier versions in public/ada9x/mrtcomments/annexes.

It'll be added to the FAQ for this newsgroup, but I wish I had more time to
read it.

>S. Tucker Taft   stt@inmet.com
>Ada 9X Mapping/Revision Team
>Intermetrics, Inc.
>Cambridge, MA  02138

Thanks for replying.

bob

Re: Ada9X, Dispatching, and Distribution (Was : Why I like C++)

[rjh]

  [Sorry if this is duplicated, its the newsreader]

In article <C7A1pD.9F3@inmet.camb.inmet.com> stt@spock.camb.inmet.com (Tucker T
aft) writes:
>>I also noted the intro document seems to assume that run-time dispatch
>>always occurs in the presence of tagged types.  
>
>It would be helpful to know what words in the document gave you
>this impression.  In Ada 9X, run-time dispatch only occurs
>...

It was really just a minor point and I see I may be wrong, a passage from
section 2.2 on message dispatch based on class wide types:

"In this case we don't know which procedure to call until runtime because
we do not know which specific type the alert belongs to."

If Process_Alerts is separately compiled, you're right since A's type isn't
known.  I *was* thinking of any of the optimizable PE cases, such as with
value propagation.  Say when:
  red_alert: High_Alert;
  ...
  Process_Alerts(red_alert);

Maybe pointing out that often the dynamic cases will run with the same
efficiency as the static cases (e.g. comparable to C; even ML does this
in a restricted sense) may help to sell the idea to those primarily
concerned run-time efficiency.

>> . . . Chamber's thesis shows
>>...
>Again, suggestions for improved wording are welcome, though we want to
>keep the introduction relatively simple.

I've been working on getting 25 hours out of every day but I've been
settling for 17 lately:-)

>The (Annotated) Ada 9X
>Reference Manual is available for those interested in all of the
>gory details.

Just got it (what a big one).  Does ANSI and ISO require all of that?
No wonder C++ is taking so long.  A co-worker asked when Ada would add
reflection too, since its the next logical step after adding OO.  He
also wondered if it would be desirable to expose all those extensive
innards, saying some wonder if Ada's really just a small language waiting
to get out.  I replied both are true, since a small well-defined reflective
kernel could be used to build Ada!  Ada 2003 maybe, how about this year?
How about a standard...

[Diana comes to mind but I don't know if anyone is using it.]

Are minor and/or major changes to the 9x standard still possible?  How will
the revisions work; will the standard be decided on at the 9x conference this
year?

>>I also would have liked to see something like Argus' non-monitor (multi-
>>threaded) objects added to Ada's active object repertoire; oh well...
>...
>implementation model, and the alternative implementation
>models could in any case be implemented efficiently 
>using the existing primitives of the langauge.

Good point.  It was just one of those experimental features attempting to
improve the tasking model.  I assume you've already covered the Concurrent-C's,
Concurrent Smalltalk's and etc. too.

>In version 2.0 of the Annotated Ada 9X Reference Manual (AARM;2.0)
>on ajpo.sei.cmu.edu (public/ada9x/mrtcomments/rm9x/v2.0/ils.aarm.ps)
>you will find a version of the Distribution Annex (Annex I).
>There are earlier versions in public/ada9x/mrtcomments/annexes.

It'll be added to the FAQ for this newsgroup, but I wish I had more time to
read it.

>S. Tucker Taft   stt@inmet.com
>Ada 9X Mapping/Revision Team
>Intermetrics, Inc.
>Cambridge, MA  02138

Thanks for replying.

bob

Re: Ada9X, Dispatching, and Distribution (Was : Why I like C++)

[agate!howland.reston.ans.net!zaphod.mps.ohio-state.edu!swrinde!network.uc]

stt@spock.camb.inmet.com (Tucker Taft) writes:
: The C++ standardization process is using the volunteer committee approach.

vs.

: Ada 9X is using a full-time design team approach with reviewers and 
: a canvass at the end.  Even so, the Ada 9X process began in October
: 1988 and we don't expect it to be fully approved until late 1994.
: The design team has been full-time since March 1990.

Clear evidence that language *DOES* influence the way people think.


--
-Matt Kennel  		mbk@inls1.ucsd.edu
-Institute for Nonlinear Science, University of California, San Diego
-*** AD: Archive for nonlinear dynamics papers & programs: FTP to
-***     lyapunov.ucsd.edu, username "anonymous".