Re: Compression/Tokens/Big Files/Small Files/and rubber ducks

Mark Waks ([email protected])
Thu, 1 Jun 95 13:01:41 EDT


Neophytos writes:
> 1) While all of this discussion may one day prove to be very important
> how important is it *today*? It seems to me that nobody has any *real*
> idea about the type of VRML files people are going to produce, because
> nobody has really generated any yet.

True, *but* it is pretty clear from history (with other VR systems) that
files are going to be Pretty Damned Big. Exactly *how* big isn't clear,
but an average of an order of magnitude larger than HTML looks likely.
Moreover, it's going to be somewhat more difficult to fetch only the
"critical" parts than it is in HTML (for example, I stop *most* HTML
uploads before it finishes the images, and start working with the page
as soon as the text is all in), although a principled use of LOD will
help this somewhat.

I see this from exactly the opposite direction you do, overall -- given
that people *already* tend to be pretty impatient with the speed of
HTML (I don't know anyone with a less-than-T1 line who isn't), we can
pretty well count upon them being even more impatient with VRML. And
when you're trying to launch a new technology, that's just plain bad.
Any *significant* improvements we can make to speed will be a Very
Good Thing -- the main argument is what is or isn't going to result
in significant improvements.

> I think Gavin's post earlier today, describing how he would create a
> VRML file of the entire planet, seems to suggest that too much of a
> fuss is being made over compression.

But that's twisting his example 90 degrees. He was demonstrating how a
file *can* be quite small (and specifically that structure-rich files
may well *be* quite small) -- but then depends on several very *large*
files. Taking the example out of context renders it nonsensical.

> What is wrong with waiting for the deployment of VRML files before we
> argue how to best compress them (*if* we decide we need this).

Simple economics. If a few VRML files get deployed, and flop miserably
because they're too slow, then there won't *be* any more files. There
won't *be* any deployment if the system is unpleasant to use. That's
simple Net psychology -- only that which people like tends to really
get anywhere. Make no mistake: we could *easily* convince the entire
Net that VRML is unusable, and lose all the momentum we've built up...

> a) a 1 meg VRML file is roughly the same as the average HTML page,

Excuse me? 1 Meg is *easily* ten times the size of an average HTML page.
I know some HTML pages that take a Meg -- I rarely use any of them.

> about 3 gifs is the average on an HTML page, and equal about 1 meg.

Nonsense -- the average GIF is around 50K. A really *big* GIF is typically
about 100K or so. Very, very few are 1/3 Meg; I can't think of any off
the top of my head.

> How many people suffer through HTML pages over SLIP? lots - with
> inline images on?

And I stop such pages before the images upload. Almost always...

-- Justin

Random Quote du Jour:

You Know You're in the SCA when...
...you're a burly guy who looks like a Hell's Angel, but you do
embroidery in public.
...you hide the really awful costume references in the stacks at
the library, so future costumers won't be led astray. Or, you write
criticisms in the margins of said awful costume references.
-- Lothar