Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!dogie.macc.wisc.edu!uwvax!rutgers!mit-eddie!uw-beaver!cornell!rochester!pt.cs.cmu.edu!wb1.cs.cmu.edu!avie
From: avie@wb1.cs.cmu.edu (Avadis Tevanian)
Newsgroups: comp.sys.next
Subject: Re: NextStep and NeWS...
Keywords: NextStep,NeWS,Postscript
Message-ID: <4792@pt.cs.cmu.edu>
Date: 21 Apr 89 09:10:11 GMT
References: <1041@nixctc.DE> <8530@polya.Stanford.EDU> <3901@ficc.uu.net> <4779@pt.cs.cmu.edu> <3931@ficc.uu.net>
Organization: NeXT, Inc.
Lines: 81

In article <3931@ficc.uu.net> peter@ficc.uu.net (Peter da Silva) writes:
>...not because sockets are too slow, but because the communication channel
>may be. I can buy an Acer Xebra 1000 for $1000 and hook it onto an ethernet
>or SLIP serial port and run X remotely. It would be more efficient to run
>NeWS on the server instead... send menu selection events instead of bitmaps.

We don't send bitmaps, we send Postscript.  I make the point about speed of
communication because it is most important when you need to do round-trip
communication between application and window server.  This simple fact
means that if you expect applications to ever need to interact with the
window server is a fast fashion (for sound, more computation, or whatever),
the speed of doing that round-trip communication is key to application
performance.  It doesn't matter if 99.9% of the application is in Postscript,
because if that .1% needs fast interaction with the window server and you
have a slow communication path, you lose.

>
>The following is a digression...
>
>> When you are doing animations (for example), do you really
>> want to write your animation in Postscript, of course not, and we don't.
>
>Hell no. I'd rather do animations in Director's scripting language, or in
>forth. Write forth programs to generate videoscape 3-d animation files.
>Which takes me back to postscript, now doesn't it?
>

Again, only if everything is in postscript.  How about adding sound
effects now?  And music synthesis in the DSP?  And...

>> What if your animation needs to be synchronized with sound effects?
>
>What if it needs to be driven by MIDI? "What if" is a trademark of HP :->.
>

Exactly my point!  You can't write everything in Postscript.  You shouldn't
have to write everything in Postscript.

>But it's still not real-time. How many messages a second could you pass through
>a pipeline containing say, a MIDI input port, a couple of filters, to a MIDI
>output port and the DSP?
>

We have done experiments with the music kit controlling synthesised
instruments running in the DSP using MIDI input and it does work in
real-time.

>Not entirely relevant, but I do think that *as yet* the advantages of Mach
>haven't fully gelled. I believe it will eventually become the final solution
>to te real-time-UNIX problem... but it's not there yet.

Glad to see you have faith in Mach.  I agree with you here, there is much
more work to be done here to fully solve the problems but we are well on
our way.

>
>Anyway, back to the point. Animation really isn't a user-interface process
>so it's kind of a straw man... I'm not talking about doing EVERYTHING in
>DP or NeWS or VideoLisp or whatever. I'm talking about a division of labor
>between the UI and the application.

Yes, and we agree, we have a layer of NextStep known as the "packages" which
is that part of NextStep that runs entirely in the window server.  It does
things like window movement, activation, ...  And as it turns out, you
as a user can load your own packages.  It doesn't allow to reimplement
the entire UI, but then again, if you do that, you don't have NextStep
anymore.  Remember, it makes little sense to reimplement window/menus/...
at the NextStep layer, because you no longer have NextStep.  If you want
to implement a different UI and allow for everything to be rewritten you
can certainly do that in DPS.


So, can we lay this to rest by saying that NextStep is a specific UI for
which it make no significant sense to reimplement features and that if
you want to have a UI in which any feature was user-implementable you
could write it in DPS?
-- 
Avadis Tevanian, Jr.    (Avie)
Chief Operating System Scientist
NeXT, Inc.
avie@cs.cmu.edu or avie@NeXT.com
--