Re: Scripts vs APIs

Linas Vepstas (linas@innerdoor.austin.ibm.com)
Wed, 7 Sep 1994 17:18:05 -0500

> To: www-vrml@wired.com
> From: mddoyle@netcom.com (Michael D. Doyle)
> Subject: Scripts vs APIs
>
> One relevant issue that the previous Viola-related discussion raises is that
> there are several different approaches to accomplishing the goals of VRML.
> Viola's approach, exploiting HTTP tranfer of and local processing of a
> client-specific scripting language, represents one end of the spectrum. Our
> approach, let's call it Mosaic-OLE for now, uses an API to allow linking and
> embedding of heterogeneous binary program objects. Each approach has
> several unique implications relating to resource requirements, computational
> load, bandwidth requirements, security considerations, etc.
>
> So far, most of the discussion here has revolved around scripting languages,
> rather than API-related approches. I would propose that API-based embedding
> would provide a more efficient means for taking advantage of the current
> state of the art in virtual reality technology as we develop the VRML standards.

There is a place on this planet for both. I don't mind scripts -- they, if
adhered to, allow VRML authoring tools and VRML visualization tools running
on multiple heterogeeous (intel, unix, etc) platforms to interoperate.

API's -- that's a different matter. We make our money selling sophisticated
3D hardware and the API's to support it. We have a vested interest in knowing
and controlling our future in the API area. VR is still a touchy enough area
that its not clear that a standards body will come up with a good solution.
Or even take the example of SGI -- they are pushing/marketing Iris Inventor
real hard, but my impression that Iris Performer is technically superior.

API's are an order of magnitude harder than scripting languages, and require
an order of magnitude more $$$ to develop, provide sample implementations,
conformance suites, bug reports, patches, tuned/optimized implementations
for different platforms, never mind things like compiler dependencies,
processor dependencies (e.g. the performance impact of passing a float on a
stack versus a float in a register ...), 32 bit vs. 64 bit growing pains.
Gag. A coder or two can hack through a scripting language in a some months
or a year. An API requires staffing departments with 10 or 20 people.

>
> **************************************************************************
> * Michael D. Doyle, Ph.D. email: miked@visembryo.ucsf.edu *
> * Director, Health Informatics Lab phone: (510)522-5275 *
> * Department of Anatomy, School of Medicine fax: (510)522-4439 *
> * University of California, San Francisco pager: (415)719-4557 *
> * 5 Remmel Court http://visembryo.ucsf.edu/ *
> * Alameda, CA 94502 alternate email: mddoyle@netcom.com *
> **************************************************************************
>

Sincerely,
Linas Vepstas

^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^
Linas Vepstas Graphics Architecture
AWD IBM Austin VNET: AUSVM6(LINAS)
Zip 9260
Dept E84S, Bldg 902 Tie Line: 678-1116
11400 Burnet Road External Phone: 1-(512)-838-1116

http://w3.austin.ibm.com/afs/austin/u3/linas/public/html/home.html
local: linas@austin.ibm.com
internet: linas@innerdoor.austin.ibm.com
^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^