Apple Dylan discussions at WWDC 
Author Message
 Apple Dylan discussions at WWDC

I spoke to a number of people at the Apple WWDC regarding Dylan's future,
including Mike Lockwood, former member of the Dylan team and now at
Metrowerks, Mike Rosetti, head of the tools group at Apple, the CEO of
Metrowerks, a couple of Apple engineers, and a guy from Digitool.
I gave out the last of the Dylan cancellation t-shirts as an effort to
raise consciousness. I tried to get the chance to talk to Patrick Beard
but wasn't able to find him after his presentation.

The upshot of all this is that the Tools group has no plan for further
work on the Apple Dylan code base. An interesting possibility raised
is that just as some of the Bedrock technology surfaced in the OpenDoc
Framework (for cross-platform development), some of the technologies
from Dylan might surface in another shape in the future.

People were pretty blunt about it: Java won. Get over it. It doesn't
matter that it isn't a better language.

It was interesting to hear the phrase "dynamic language" mentioned
during the Java tools presentations. One presentor showed pasting a
fragment of Java source code into an inspector-like window and evaluating
it to create a functional AWT window. "Almost like a dynamic language"
was the phrase used. Someone on the Roaster team also talked about
being able to rewrite a Java class on the fly, but it really isn't the
same thing that a fully dynamic language like Dylan or NewtonScript can
give you.

I had the chance to try building and running with Digitool's alpha
Dylan TR and it does indeed build and run much faster, although there
are appeared to be a number of bizarre display-updating bugs; Apple
Dylan's beautiful browser interface has always had strange redrawing
problems. Digitool doesn't have plans to do much more than make a
PowerPC-native version of the Technology Release.

Someone indicated that a surprisingly large number of copies of the
TR had been sold. If this is a true, I wonder if there is a latent
market for a good Apple Dylan. (This "good" Apple Dylan would probably
have to be a reimplementation of Dylan in Dylan, not built on Mac Common
Lisp; I don't know exactly why it is too slow, but it is indeed too slow;
it seems to me that with RAM finally getting cheaper, the world might
be ready for a very big development environment for a dynamic language,
but if it is painfully slow, it just won't sell).

Now that Java is a buzzword de rigeur, a major announcement was (of course)
that Apple will start supporting Java everywhere, (of course) whether or
not it makes sense, such as in the Newton OS, even though NewtonScript was
designed from first principles to be extremely conservative in its memory
usage, and Java wasn't.

I don't know anything about the status of Harlequin's Dylan development
environment for Windows, but the irony of a Windows version shipping
with no competing Mac environment available would not be sweet.

-Paul-

--
Paul R. Potts - Technical Lead - Health Media Research Lab
University of Michigan - Comprehensive Cancer Center



Fri, 06 Nov 1998 03:00:00 GMT  
 Apple Dylan discussions at WWDC



Quote:
>...
>People were pretty blunt about it: Java won. Get over it. It doesn't
>matter that it isn't a better language.
>...

Sigh.

If that is their attitude, I wonder if they would then be willing to release
the source, maybe on the "update" CD for the native version...

-Robert
--

      "The intellectual level of the schools can be no higher than the
       intellectual level of the culture in which they float."
                                                     -Richard Gibboney



Fri, 06 Nov 1998 03:00:00 GMT  
 Apple Dylan discussions at WWDC



...

Quote:
>People were pretty blunt about it: Java won. Get over it. It doesn't
>matter that it isn't a better language.

...

  Pretty ironic that folks at a Macintosh conference would have this
  sort of attitude.  Simple text substition of a few words above is:

 "People were pretty blunt about it: Windows won. Get over it. It doesn't
  matter that it isn't a better OS/Environment."

 They could use the same "logic" and told everyone who came to go off and
 worship those almighty in Redmond....

 It is not an exact analogy, but for a company who's current slogan is
 "Expect the Impossible", they certainly "frighten" pretty easily.
 There need not be one and only one "hammer" for developers to use.
 I thought Apple wanted to aid developers in search of quixotic applications
 to build.  If a building a competitive Dylan environment isn't
 quixotic, what is? :-)

 Yes, given the hype that Java is receiving, Dylan will never
 dominate.  But that doesn't mean it can't "play".

Quote:
>I had the chance to try building and running with Digitool's alpha
>Dylan TR and it does indeed build and run much faster, although there
>are appeared to be a number of bizarre display-updating bugs;

  Hopefully, those will disappear in the Digitool port. I think the
  editing windows were build on top of FRED (which changed from MCL 2.X
  to MCL 3.X and probably with significant ramifications on the code).

Quote:
>Someone indicated that a surprisingly large number of copies of the
>TR had been sold. If this is a true, I wonder if there is a latent
>market for a good Apple Dylan. (This "good" Apple Dylan would probably

   The market dynamics of a $39.00 product are different than a $390.00
   product.  Not that it couldn't be done, but how fast those large numbers
   shink as you proceed down the "demand curve" from $39.00 to $390.00 is
   what mainstream toolmakers aren't so happy about ( or perhaps not too
   sure about).

Quote:
>Lisp; I don't know exactly why it is too slow, but it is indeed too slow;

  If slow means the browser...

   How about the non-native file system?  There is an OODB sitting down
   below the browser et. al.  Depending upon what sort of caching scheme
   ( which likely isn't "tuned" ) is implemented that could have a
   distinct affect on performance.  Not to mention interaction effects
   between said caching scheme and Apple's so called Virutal Memory system.

   Note that Digitools benchmarks for the  MCL-PPC hover around 3-times MCL-68k
   EXCEPT when doing alot of file I/O, then it was something like 1.5 times
   improvement.

   Additionally, I would beat that with some profiling info a really good
   Lisp programmer could get some noticible speed improvements. As the
   Environment exists now I think the primary obejctive was to get the  
   thing working. Then worry more about speed ups.

  If slow means the resulting Dylan code...

   Compiling Dynamic code into nearly-fast-as-static code will require a
   seriously non trival optimizer which probably will gobble RAM and/or time
   in such a way as to make the current environment look tame. However, you
   need only do that when delivering a product not neccessarily in development.

--

Lyman S. Taylor              "We are Microsoft.  

                                  resistance is futile.  
                                        Prepare to be assimilated."
                                .sig seen while netsurfing the Internet.



Fri, 06 Nov 1998 03:00:00 GMT  
 Apple Dylan discussions at WWDC


:  They could use the same "logic" and told everyone who came to go off and
:  worship those almighty in Redmond....

First, let me say that I agree with your arguments: Java isn't superior,
and it's frustrating to see Apple tossing off Dylan.

OTOH: the Java:Dylan relationship bears some striking resemblance to the
current Windows:Mac one. Also, Apple lost $700+ million last quarter.
Maybe in a few years...

Personally, what I would love to see is a Dylan-like development environ-
ment for Java. My first thought upon booting up the Apple Dylan DR was,
"Wow! We shoulda had this *years* ago for C/C++." Then I realized you
can't, due to the preprocessor, dependence on files, and so forth.
But Java seems suited to this much better; as a matter of fact, the
requirements on file structure (one class per file, filename must match
class name, directory structures, etc.) seem like Gosling was trying
to match a database as best he could given reasonable constraints.

I've lost my Mac religion, I'll admit. While I used to be a fan of the
Mac and Dylan, now I'm moving to Linux/WinNT (the only even reasonably
tolerable Windows) with C++/Java for actual work and the BeBox with
C++/Java as my religious platform. At least, that's where I expect to
be in a couple years unless Amelio gives a miracle.

The truth of the matter is that while platforms can survive with 1%
of the market, languages can't have that kind of marketshare and have
development tools good enough to produce shippable apps. It's very
frustrating.

                                                        / ag



Sat, 07 Nov 1998 03:00:00 GMT  
 Apple Dylan discussions at WWDC


Quote:

> If that is their attitude, I wonder if they would then be willing to release
> the source, maybe on the "update" CD for the native version...

This was also a topic of discussion with some Apple engineers who shall
remain nameless: the possibility of releasing the source under the Gnu
public license.

This could be problematic because there are software patents associated
with the Apple Dylan implementation. I don't really know how such a thing
could work, if it could.

I don't see a good analogy between Dylan/Java and Mac/Windows. Sun has
opened up the Java spec and is giving away a reasonable implementation.
Apple published the language spec and is selling a problematic
implementation which us undergoing no further development.

Java will lend itself to somewhat more dynamic implentations and I
suppose it would be possible to create a version that had some more
support for dynamism, but the current spec is lacking in that regard.

Sun was looking at proposed extensions to the Java language - as far
as I'm concerned, support for a more dynamic method of programming is
more of a selling point than, say, operator overloading.

-Paul-

--
Paul R. Potts - Technical Lead - Health Media Research Lab
University of Michigan - Comprehensive Cancer Center



Sat, 07 Nov 1998 03:00:00 GMT  
 Apple Dylan discussions at WWDC


Quote:

>   If slow means the browser...

The browser was slow to update but I can easily believe that the built-in
object database is slow as well, and certainly the Mac's file system doesn't
help much. (and, of course, VM in existing Mac implementations performs quite
poorly; an environment such as Windows NT has far superior virtual memory
performance. I have high hopes for MacOS 8).

Quote:
>    Additionally, I would beat that with some profiling info a really good
>    Lisp programmer could get some noticible speed improvements. As the
>    Environment exists now I think the primary obejctive was to get the  
>    thing working. Then worry more about speed ups.

Exactly... this was a good way to prioritize, the problem is that there is now
no one left to work on the speed tweaks.

Quote:
>   If slow means the resulting Dylan code...

>    Compiling Dynamic code into nearly-fast-as-static code will require a
>    seriously non trival optimizer which probably will gobble RAM and/or time
>    in such a way as to make the current environment look tame. However, you
>    need only do that when delivering a product not neccessarily in

development.

Actually, the experience I have with the alpha and TR1 (limited though it is)
has convinced me that the compiled code is actually *quite* fast. Another reason
that I think a real Apple Dylan implementation needs to be built in Apple Dylan.
After all, any decent C or C++ compiler can build itself.

-Paul-

--
Paul R. Potts - Technical Lead - Health Media Research Lab
University of Michigan - Comprehensive Cancer Center



Sat, 07 Nov 1998 03:00:00 GMT  
 Apple Dylan discussions at WWDC

Maybe its kismet, but the following interview, with none other than
Gosling himself, was just published in _Java Report_. [May/June 96,
published by SIGS -- www.sigs.com]

"""""
Q: Does this mean the Java is, you know, essentially just a subset of C++?

A: Not So. We're looking into what it takes to support other languages on
top of the Java Virtual Machine. And among those are things like Visual
Basic, and some of the list languages like Dylan [Woohoo!!] and
Scheme...

He goes on to talk about a compiler for Ada that writes Java byetcodes,
and then says:

...The Java language comes from a C++ tradition. But one can easily
imagine some people being more effective using languages from a list
tradition.
"""""""

This is very encouraging. I don't know about anyone else, but I'm actually
starting to feel pretty good about our prospects for escaping the dark
ages of C++ {*filter*}. As an aside, I have spent a lot of time with Java
in the last few months, and while it is far from perfect, it is actually a
pretty satisfying and entertaining solution, until the perfect answer
arives. I know I'm hardly the first person to point this out, but I think
this may have been a big part of Dylan's (and Apple's) problem; At some
point the perfect became the enemy of the good.

As far as using the Dylan IDE for Java goes, I just wanted to mention that
there are some aspects of the 'Binder' and, to some extent, the class
browser, that aren't as appropriate for a less wonderful :-) language like
Java. Specifically, methods belong to classes in Java, so you'll get far
less granualarity and flexibility in the binder. Of course, it _would_
still be really neat to be able to have an editor that supported
sumamrizing and expanding method definitions within classses. Why someone
hasn't allready done this with a commercial C++/Java environemnt, I have
no idea. Also, the object model is less ubiqutous in Java,
IMLimitedExperience. I think a lot of the Dylan browser stuff relies on
having objects from the ground up, but I could be way off base here.

-miles

_______________________________________
Miles T. Parker          Washington, DC
Database/Macintosh/Internet Development

_______________________________________



Sat, 07 Nov 1998 03:00:00 GMT  
 Apple Dylan discussions at WWDC



Quote:
>  Pretty ironic that folks at a Macintosh conference would have this
>  sort of attitude.  Simple text substition of a few words above is:
> "People were pretty blunt about it: Windows won. Get over it. It doesn't
>  matter that it isn't a better OS/Environment."

While the analogy is proper, it's easy to understand why they don't take
that attitude.  The Macintosh and MacOS are Apple's core technology -- if
they abandon it, they might as well go out of business.

Dylan, on the other hand, was a new project, and not critical to the
company's success.  It would have been nice if it had succeeded, but they
didn't have so much sunk into it that they couldn't punt and switch to
Java.
--
Barry Margolin
BBN PlaNET, Cambridge, MA

Phone (800) 632-7638 - Fax (617) 873-6351



Sun, 08 Nov 1998 03:00:00 GMT  
 Apple Dylan discussions at WWDC

Quote:
> As far as using the Dylan IDE for Java goes, I just wanted to mention
that
> there are some aspects of the 'Binder' and, to some extent, the class
> browser, that aren't as appropriate for a less wonderful :-) language
like
> Java. Specifically, methods belong to classes in Java, so you'll get
far
> less granualarity and flexibility in the binder. Of course, it
_would_
> still be really neat to be able to have an editor that supported
> sumamrizing and expanding method definitions within classses. Why
someone
> hasn't allready done this with a commercial C++/Java environemnt, I
have
> no idea. Also, the object model is less ubiqutous in Java,
> IMLimitedExperience. I think a lot of the Dylan browser stuff relies
on
> having objects from the ground up, but I could be way off base here.

FWIW, Java WorkShop does something similar to this (if I understand you
correctly) - it generates on-the-fly Javadoc for classes (ie method
definitions) and a click takes you into the source code (but in a
separate window, whereas I assume with the Dylan IDE you get it in the
same window).
BTW, note that the Dylan spec is quoted as one of the standard
references at the start of the latest Java Language Spec....

Matt.
--

Words, thoughts and opinions all my own, after all....
"I wouldn't say it if I didn't mean it." ('Lost Weekend' -Lloyd Cole)



Sun, 08 Nov 1998 03:00:00 GMT  
 Apple Dylan discussions at WWDC


:  Q: Does this mean the Java is, you know, essentially just a subset of C++?
:  
:  A: Not So. We're looking into what it takes to support other languages on
:  top of the Java Virtual Machine. And among those are things like Visual
:  Basic, and some of the list languages like Dylan [Woohoo!!] and
:  Scheme...
Well Gosling is certainly aware of Lisp.  Back in his graduate student days he wrote the precursor to GNU Emacs which had a Lisp-like extension language (excect, if I remember correctly, it didn't CONS).
-David



Sun, 08 Nov 1998 03:00:00 GMT  
 
 [ 64 post ]  Go to page: [1] [2] [3] [4] [5]

 Relevant Pages 

1. Dylan at Apple WWDC

2. Dylan at Apple WWDC Report?

3. Oberon microsystems at Apple's WWDC

4. Smalltalk at Apple WWDC? (QKS responds)

5. Smalltalk at Apple WWDC?

6. Dylan Programming Book and Apple Dylan

7. Dylan is the Name was(Re: Dylan (Bob) eats rotten Apple (Computer))

8. Dylan WWDC video tapes?

9. Dylan at WWDC

10. Old Apple Dylan Archive

11. Apple Dylan TR

12. apple's dylan technology release CD

 

 
Powered by phpBB® Forum Software