Alex Bowden wrote:
> Justin
>
> Your long justification sounds too much like a politician and almost totally
> misses the point based on the interaction

First time I've ever been accused of sounding like a politician! I think
I'll have to go back to swearing every second word then :)

If I've missed the point, its because there hasn't been one to miss. I'm
answering straight up your list of issues raised. Namely: what's our
responsibilities - percieved and real, what are we doing with the code,
and quite clearly stating that "VRML97.jar" has nothing to do with Us,
the Web3D Consortium or anyone else except for Sun. Take the issue up
with them.

> I cannot see how you can possibly claim that XJ3D is "better" and that "The
> old loader is redundant and has been replaced by a newer system."

Back in circles again - your definition of better is different to ours.
Here's a simple list of why we think it is better: More nodes supported,
  spec compliant, more flexible (everything replacable and extendable),
new and old specs supported, current, active development. For a single,
simple comparison - try adding a Material node to an Appearance in Sun's
code after the scene graph has been loaded.

> They are different and until the speeds become comparable there will be a
> need for both types of facility.  You indicate (which suprised me) that you
> expect that situation to continue.

Why should it surprise? What else were you expecting? "Yes our 200K LoC
project that supports 10 times the amount of features will be as fast as
one that barely even recognises the word VRML"? (FWIW Sun's loader
doesn't even check the file header for compliance).

> Nobody is suggesting the XJ3D should support the original loader -

Strange, that seems to be your entire point of the thread.

> Meanwhile, the speed problem has been consistant back from M2 through M4 and
> isnt particular to the latest build so I have doubts about the
> IndexedFaceSets idea.

We already have a bunch of test files and user reports about this, so we
know there is definitely an issue that cropped up there. We'll fix that
and then get back to you for more testing. There's been another huge
internal shakeup of the code, and in particular the node loading process
  now that the X3D profile/component spec has settled into a final form.
It's going to be hard to do direct comparisons for profiling of one
piece of code from the M4 code to current, so we want to hit one issue
at a time.

> It there is any real interest in performance issues then I am perfectly
> happy to send whoever a data file provided that it is treated as
> confidential (the larger data files around are all client data and not
> mine)

The performance issues we are currently targeting are runtime, not load
time. Even then, performance is less important for us that completing
support for the rest of the spec. Our current contracts are paying for
completion and spec compliance/testing, not performance. If you
want/need performance fixed in the short to medium term, then money can
always lubricate those squeaking wheels or you can fix them yourself and
send us the patches. :)

We already have a bunch of test files with similar conditions on them.
We'll get back to you later about this. If we forget, can you send it to
us at the point we annouce the first test release for M5? Should be
around late June/early July (we're targeting Siggraph). Sending it
earlier and it will probably get lost in all our other things that are
currently happening.

Hmm.... just a random thought..... Did this start with the M2 code or
later? The Sun code had a patch that was run on the JavaCC generated
source because of a bug that was not nulling out the Token data
structure chain references. This would cause really large files like CAD
data to not have these structures GC'd and would heavily impact
performance. We never used that in the latest versions of code because
we assumed that JavaCC 2.1 had fixed that problem. Maybe something else
to look at.....
--
Justin Couch                         http://www.vlc.com.au/~justin/
Java Architect & Bit Twiddler              http://www.yumetech.com/
Author, Java 3D FAQ Maintainer                  http://www.j3d.org/
-------------------------------------------------------------------
"Humanism is dead. Animals think, feel; so do machines now.
Neither man nor woman is the measure of all things. Every organism
processes data according to its domain, its environment; you, with
all your brains, would be useless in a mouse's universe..."
                                               - Greg Bear, Slant
-------------------------------------------------------------------

===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST".  For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".

Reply via email to