Models are not really comparable to USD.  Models are more similar to 
Houdini's concept of Digital Asset than a scene.

The purpose of a model was to be a container for an asset so it could be 
more easily included in large scenes and independently manipulated with 
regards to versioning and level of detail control.  The main message of 
'Sumatra' and DS was that artists were to be removed from the burden of low 
level manipulation, such as setting key frames, and able to focus more on 
higher levels of control such as choreography via animation mixing.

XSI Models are derived from the "hierarchy" (.hrc) from predecessor 
Softimage|3D.  One could argue the Model is a superset of the .hrc as the 
main difference is the addition of the animation mixer (independent 
timeline).  The .hrc had level of detail control and other mechanisms, but 
they were rarely used due to lack of tools to do it in a practical manner by 
the end user.  The Softimage|3D implementation required a lot of digging 
through menus and strong self-discipline in organization of files.  For some 
operations, one had to jump out to the command line and use DbTools - a 
proprietary environment for running shell scripts developed by Softimage.

Softimage|3D had an ASCII text scene file format which related multiple 
files, this was closer to the concept of USD.  The main file was the Scene 
Description (.dsc) which contained all the high level information of what 
elements were in the scene and how they were related.  This included 
relations such as parent/child, master/slave relation of constraints, light 
and model associations, groupings, custom effects and their parameter 
settings, asset LOD, and the list goes on.  It even includes position of 
nodes as represented in the schematic view,  The Scene Description was 
accompanied by the Scene Setup file (.sts) which contained description of 
the user's environment such as how the viewports were arranged, current 
settings, render parameters, etc.  All the other files pointed by the Scene 
Description and Scene Setup were binary files and contained the actual 
content (cameras, lights, models (hierarchies), materials, animation, ...). 
This meant a scene was a conglomerate of many files - handy for versioning 
and substitutions, but took forever to load/save when the database started 
to fill up.  A single scene could consist of thousands of files, and each 
file would have to be checked individually for version collision, and sync 
with the other assets in the database.  If scene save operation was 
interrupted (e.g. crash), you had many orphaned files floating around in 
your database creating all sorts of havoc.  Since the default number of 
versions kept was 4, and all hard drives of the day were disc drives with 
small caches and slow spin rates and 10 Mbps ethernet (if lucky), it took 
forever to get work done.  That was a strong motivation behind XSI using a 
single file for it's scene file format, and the .scntoc to mimic (to a much 
lesser degree) the Scene Description.  Ironically, Alias|Wavefront did the 
exact opposite as Power Animator used a single file scene format, but 
migrated to the Softimage|3D style with Maya.

USD is more robust in that it's a rule based language to do the work of 
assembly and has some amount of intelligence to handle collisions and other 
problems.  XSI models had limited ability to do that in the delta, but it 
was self contained.  XSI Scenes had some back door exceptions to the rule to 
accommodate models in certain situations, but otherwise were passive and not 
available to the user to manipulate.  The Softimage|3D scene description 
language was more primitive than USD in that it was more like a passive list 
of commands to execute than a language with decision making intelligence. 
Although error checking existed, it was largely assumed the contents of the 
file were ordered and compliant.  If a model was listed twice, for example, 
the parser would not catch it and made every effort to import the model and 
establish all the relations.  It would be left to the other operations 
downstream to catch the problem and report it.

USD isn't rocket science, but it serves a very important role that's been 
long needed.

Matt



Date: Tue, 27 Aug 2019 10:13:05 +0100
From: Jordi Bares <[email protected]>
Subject: Re: Thoughts on USD in comparison to Softimage 'Referenced Models'
To: "Official Softimage Users Mailing List.

I guess the concepts are rooted in our reality and therefore it is only 
natural the approaches are not that dissimilar, but it is fun to realise 
Softimage Models were so advanced at the time as to be compared and all to 
USD.

IMHO USD is a major leap forward and models and references are a minor 
attribute as the real meat is the whole versioning in combination with 
layering which somehow makes me thing of XML document conversions for the 
web and data mining but I digress..

USD is a framework and the scope is enormous, to the point that you don?t 
need Alembic if you don?t want to, and chances are materialX becomes 
irrelevant as well. I see the Houdini implementation presented at Siggraph 
and it is in my little knowledge a true disruptor to the point that if it 
will leverage scale to the masses, something we don?t have yet so I expect 
companies to start exploring it and moving towards it real fast.

Furthermore, Apple is amongst many big companies investing tons of money 
into it as they have decided it to be their standard in the iOS and OSX 
content system, for all VR expect to see a lot of USDZ (the compressed 
version) and as such, I feel the industry money being pouring into USD will 
become very obvious very soon.

Then you add Hydra to the mix and the whole concept of a render engine 
becomes quite more subtle as we will see the real differences between them 
in how they implement algorithms? right now there are many other important 
tasks the render engine does that via Hydra it won?t (think the deep 
integration of Mental Ray with Softimage but in a generalised way?) Meaning, 
we are going to see speed increases on the pure raytracing more than texture 
management so I hope that allows these engines to focus on what is important 
and for us artists to move to the engine because we need those features.

IMHO this is a whole new world bigger than reference model, this is proper 
asset management and rendering pipeline all in one.

My 2 cents jb


------
Softimage Mailing List.
To unsubscribe, send a mail to [email protected] with 
"unsubscribe" in the subject, and reply to confirm.

Reply via email to