1  2  3  4  5  6  ...  13  |   Order by default

Jul 16, 2003, 10:43am

Hear hear!

[View Quote]

Jul 16, 2003, 12:36pm

I would like some more detailed info on how exactly the browser
determines if content is stale for a given object path.

The docs imply that the browser 'checks the date', but the only
way to do so is by looking at the HTTP headers, and there are
a lot of different kinds of HTTP headers that control content caching.

I am noticing that any content created via 'create picture' commands
is downloaded completely anew every session even though there
is no 'update=' attached to it, and the server is reporting the content
as fully cacheable.

I have a museum with about 60 nice hi-res monet images that are
completely static, but every time someone comes into the world
they have to fully download about 4 megabytes of images because
they are all implemented via a 'create picture'.

I know I can workaround this using create texture instead of create
picture (assuming I construct a picture object that is to my liking and
does not use any textures for any other parts of itself), but it would
be nice if 'create picture' just honored the servers caching policies
and simply checked to see if the file has actually changed first.

Jul 16, 2003, 2:07pm

Whether the browser is using mipmapping or not has nothing to
do with my question. Mipmapping is something the browser may
or may not do to my content once it has downloaded it, but it
certainly has no effect on whether the content has actually
changed, or on whether the content should be downloaded
again if it has not changed.

And there is no reason that the mipmaps should change if
the content has not. If I give you an image today, the mipmaps
you generate from it are going to be the same as what you
generate tomorrow if I give you the same image.

If you connect to your ISP through a caching proxy server
like most large ISP's *they* will honor the caching headers
I am putting on my content, and all the AW browser is doing
is downloading the content from your ISP's hard drive. They
aren't getting it again from me.... so mipmaps or not.. the
browser is wasting bandwidth needlessly.

I can prevent this from wasting my servers bandwidth to some
extent, but I would like to also prevent the users from having
to download content when they don't need to. If I want the
users to update the images used with create picture, I will put
an explicit update command on only those items.

What I would like is some official guidance from AW about
how to work with their browser to conserve bandwidth and
ensure a better user experience.

Maybe (in the spirit of hacks) they could allow update=-1
which would mean 'never'.

[View Quote]

Jul 16, 2003, 3:22pm

That's great news. So all that is needed is for someone to provide
the detailed information I asked for in my original post, and everything
will be just peachy.

To recap:

"What is the specific method the browser uses to determine if
content is out of date and needs to be downloaded afresh?"

I already know it "checks the date". Which date does it check? Does it
check the "Last Modified" date, the "Expires" date or does it look
for an ETag? does it use the Cache-Control header max-age value at
all? Does it honor any of the other Cache-control headers?

Knowing that your world works doesn't answer my question any
more than knowing that pictures don't use mipmaps.

[View Quote]

Jul 21, 2003, 10:51am

>More
> likely it's the exact environment that's needed to produce the
> reflection duplicated somewhere else on the map that's got a camera in
> it, and when characters come near it, they're drawn into the reversed
> environment as well (reversed too). Something along those lines.
> Certainly the easy to way to do it.

There's no need to duplicate the environment when the environment
is already there.

Generally, a camera is simply placed exactly where the mirror is, a small
image rendered from that viewpoint every once in a while, then it's flipped
left to right by just mapping the UV coords so the picture comes out
reversed.

The camera takes one image for each person who sees it, in a direction
that is away from the perpendicular in the opposite direction to the users
viewpoint, so that it looks like you are seeing things just as a mirror
would
show them. It's very similar to how a 'facer' works... always pointing
towards you, but in this case it is always pointing away from you by
as much as you are not facing the mirror directly.

Jul 21, 2003, 9:45pm

There is a good article at:

http://www.sgi.com/software/opengl/advanced96/node28.html

on how to create a mirror effect with a 3D graphics engine.

As you can see from this article, a stencil buffer is merely a way to mask
out the mirror portion of the main scene so that something else (like a
mirror) can be drawn into it without interfering with other parts of the
scene.

A stencil buffer in no way prevents you from having to render the scene
twice as you imply.

There is no way to get around having to draw the scene from the viewpoint
of each mirror you wish to have in your scene, but since the mirrors could
be limited to using a 128x128 texture and they do not have to be updated
as often as the main image, there is no reason that it has to create lag.

It is also very easy to detect (since a mirrors position and field of view
are
fixed) when the mirrors image doesn't need to be updated at all.

Mirrors reflecting other mirrors handle themselves automatically. after one
or two frames, each mirror will naturally display the last frame displayed
by the other mirrors.

[View Quote]

Jul 21, 2003, 9:54pm

A more accurate description of what I was trying to say

--

The basic steps to calculate the image from the vantage point of the mirror
are:


1.. draw the mirror polygon, setting the stencil buffer to 1.
2.. compute the angle between the viewer and the mirror.
3.. compute the angle of the reflected ray and the viewer
4.. move the viewing position to a point along the reflected ray
5.. draw the scene from the viewpoint along the reflected ray
6.. set the stencil test function to draw where the stencil value is not
equal to 1
7.. move the viewpoint back to the eye position and draw the scene again
> The camera takes one image for each person who sees it, in a direction
> that is away from the perpendicular in the opposite direction to the users
> viewpoint, so that it looks like you are seeing things just as a mirror
> would
> show them. It's very similar to how a 'facer' works... always pointing
> towards you, but in this case it is always pointing away from you by
> as much as you are not facing the mirror directly.
>
>

Jul 21, 2003, 10:21pm

My apologies.

It's a description of a procedure.. not a .exe file

And don't type it in and expect it to work either. It's not real source
code.

[View Quote]

Jul 21, 2003, 10:28pm

OK.. welll I suppose I should have cited the source.. but you can
go tell them their procedure "might not work"...

http://www.sgi.com/software/opengl/advanced96/node28.html

They are all n00bs there... go easy on em Bowen.. I shudder
when I think of how they will tremble at your wrath.

[View Quote]

Jul 21, 2003, 10:52pm

The main point here is that if you want mirrors you need to render the
scene multiple times.

KAH also stated that stencil buffers would magically allow for reflections
without drawing the scene a second time, but they won't.

There are lots of ways to get around not having a "stencil buffer", but
since a stencil buffer (or equivalent) is an attribute of the graphics API
you are using and not of your hardware it is up to the driver to simulate
it if you have lame hardware that doesn't have the ability to make
them work more efficiently.

In DirectX you don't need stencil buffers because you can have as
many buffers as you wish, any buffer can be combined with any other
buffer in ways that make stencil buffers look like Flintstones technology,
and in fact you don't even need to use a stencil since you can simply
render into one buffer, and then use it as a texture map when you
draw the mirror.

So the "no card no stencil buffer" thing is a RED HERRING

So.. reflections require multiple scene renders with or without a
stencil buffer.

And.. if you are willing to draw the scene a second time from
the viewpoint of the mirror, you can have a nice mirror whether
the person has stencil buffers or not.

[View Quote]

Jul 21, 2003, 11:10pm

The nasty thing about reflections is not how to render them, but what they
would do to the current AW 'one viewpoint' model.

Right now, the engine has developed over a long time to be efficient about
what objects and textures to load based on the assumption that it only needs
to know what is near the viewer.

Introducing reflections breaks that assumption in a big way, and I suspect
that assumption is entrenched pretty deeply in the code.

Maybe an easy way to get around that is to limit what a mirror sees to a
fixed distance, and then make sure you always have that much 'stuff' from
in front of the mirror at all times. Putting many mirrors around your world
could lead to a lot of problems though.

Most game engines don't need to worry much about that because they
load all of the geometry and textures for a given area ahead of time.

[View Quote]

Jul 22, 2003, 12:09am

> I guess, what I was trying to get at, was something that could be done
> regardless of the API you're using. Some graphics cards don't even
> support OpenGL.. don't forget about them either.. if they do it's not
> too well.

The technique of drawing the scene from multiple viewpoints to generate
reflections *is* something that can be done regardless of the API you
are using, and the stencil buffers, lack of them, presence of them,
use of OpenGL vs. Direct X are all RED HERRINGS.

And as Anne so eloquently said in another thread, I'm sick of being held
back by people with lame hardware. If your card doesn't support OpenGL
or DirectX, then AW is not for you.

Now I'm finished belabouring the point with you... you can have the last
word.

Jul 21, 2003, 11:29pm

Given that people are asking for reflections, I would like to ask
(I remember reading someone else asking for them) for cameras
and displays.

A camera could be created on a given object like so:

create camera "name" [other parameters]

It would nominally point in the direction of one of the axes of
the object like a spotlight.

A display could be created like so:

create display source=cameraname [other parameters]

It would probably need to be a picture object, or some new tag.

The other parameters could be things like update rate, FOV for
a camera, tracking options to automatically pan the camera around
etc.

I think such a mechanism might actually work to make a pseudo
mirror, although it would not be a true mirror as discussed in
another thread.

If there were an option to make a camera automatically face away
from the viewer by as much as the viewer is positioned away from
the surface normal of the camera though, it should actualy work
perfectly.

create camera,display source=me pan=mirror_viewer

Something like that would create a mirror.

If this pair of commands could simulate mirrors, it could also
provide for live in world cameras too.

Jul 31, 2003, 11:59am

Adobe Atmosphere does avatars this way. They don't need to be on the
users computer though... they can be anywhere on the internet.

The main point being the persons avatar is not determined by the world
they are in, but by their own personal choice.

[View Quote]

Aug 4, 2003, 11:48am

Is there a 'reserved' cit number that can be used to 'lock' objects using
a bot that we can count on never being used as a real cit number?

i.e. "1" is probably a good guess... but it *may* actually be used.

I'd like to ask for a 'reserved' cit number for the purposes of marking
objects as 'infrastructure' so that it will be easier to use bots to filter
things etc.

It would be like marking files as owned by 'root' in UNIX.

It would be nice if such items could not be deleted by anyone but the
world owner, but that anyone (even tourists if you enable them to build)
could build in cells that have items owned by 'root'. This would allow you
to make public areas that have infrastructure that can't be changed, but
that anyone can still build on.

I'd settle for just a reserved cit number though...

Aug 4, 2003, 12:48pm

What do you mean a 'world' cit?

The main reason I want this is for using bots to filter things..
not for the purposes of restricting building.

I want a cit number that will *never* be a valid citizen.

If I use a cit number of someone who actually might build in
a world, then I cannot filter that persons objects using a bot.

I do this now using a cit number I don't *think* will ever be
used by anyone in my world, but I would like a sure thing.

[View Quote]

Aug 6, 2003, 12:17pm

Some programmer once said... geez.. two digits is plenty to use
for the date.

Another programmer once said if I use the number of seconds since
Jan. 1 1970 as the time, it will be good forever. In 2038 we will find
out how long forever is.

And of course AW is just about to reveal their new marketing plan
and the number of cits a month registered will shoot up to 10,000

[View Quote]

Aug 7, 2003, 6:04pm

You're assuming that the little computer in the gas pump will be
replaced with a new one by then and that somone will actually
remember that it uses a 32 bit processor, or that it even has a
processor in it.

Servers are only one little tiny application for computers, and
embedded computers.. well.. they are *embedded* in things.
They don't get routinely swapped out, and they are not designed
to go obsolete in a year either.

It's likely that most embedded systems will have been replaced
by then, but who knows.

[View Quote]

Aug 7, 2003, 11:00pm

I didn't mention payment of any kind in my post.

> You don't pay gas with cash? That's a bad practice, using a credit of
> sorts for gas. :P

Aug 8, 2003, 12:07am

A UNIX timestamp is a UNIX timestamp wether implemented on an
eight, 16, 32 or 64 bit computer.

And it is the same wether you pay by cash or by credit card.

You're an expert at missing the point.

[View Quote]

Aug 8, 2003, 4:18am

I'm ignoring your point, not evading it.

You made it up to imply that I said anything about cash to begin
with. You have this notion that keeping track of anything having to
do with timestamps involves cash somehow and it's totally missing
the point.

And I don't post in the newsgroups to engage in long winded pointless
debates with you. Find someone else to annoy.

Aug 8, 2003, 6:24pm

*this* is your original *question*.

"You don't pay gas with cash? That's a bad practice, using a credit of
sorts for gas. :P"

It is *not* a question.. it is an insult... flame bait.

It is worth ignoring, even if you try to change it into a question after the
fact.

You may become a veteran newsgroup poster someday, but you
currently lack the subtlety and refinement in your posts needed to
disguise the bait properly. Work on your non-sequiturs.. they are way
too blatant.

Aug 4, 2003, 1:43pm

It would be nice if there were a way for text signs to mask themselves.

A special bcolor= command could be used to do this.

Aug 10, 2003, 3:44pm

This could all be done now on the OP path server by third parties, but the
caching mechanism of the avatars.dat file and the avatars themselves by the
browser really gets in the way of developers doing anything useful with
anything like that.

[View Quote]

Aug 14, 2003, 10:53am

[View Quote] Every time I have to manually zip a mask for an image, I curse the lack of
support for a format with an embedded mask.

The fact is that it's a pain in the ass to have to create a mask seperately
as
a .bmp and then zip it with a different name while the .jpg that goes with
it
is not zipped.

Most companies spend a lot of effort trying to make it easier for new
users to use the product. An extra meg of code in the executable is
well spent if it saves users time.

How much is it worth to save one minute of time every time every world
owner wants to create a masked image? Forget the extra minutes that will
be saved correcting the mistakes made in one of the 5 or 6 things you have
to get right to create a masked image now. 2,000 worlds.. 5 times a year
apiece.. that's 10,000 minutes a year right there.

Personally, I'd eat the extra meg in the browser to have .png support

> Unless gifs are animated, they suck compared to JPGs...

... or unless you use the fact that they have a transparent color.. or the
fact
that a pallette based gif is much, much much smaller than a .jpg when you
don't need photo quality images.

> .tga is a worthless format only used in some games.

Yeah.. it's pretty old in the tooth.. but it does support a mask, and it's
dead easy to support that format. It compresses well in .zip files too.
One or 2K of normal code, 10 or 20K of windows code to decode
a targa file.

Aug 14, 2003, 8:08am

It's already there. Look in your new control configuration panel and map the
keys to somethiung.

[View Quote]

Aug 14, 2003, 5:52pm

[View Quote] I notice you're setting those keys up on keys that already have other
functions mapped to them.

A really good feature of the new key configuration thingie is that the
various Numpad keys don't conflict with the insert/home/Pgup/pgdn
keys in that other block of six keys on a 101 key keyboard.

The default mapping maps both the numeric keypad keys, and the
UN-NUM-Locked equivalent keys to the same functions.

But if you have a 101 key keyboard, you can treat that block of six
keys (ins/del/home/end/pgup/pgdn) that is not on the numeric keypad
as pretty much new keys entirely, continue to use the numeric keypad
for things you've always used it for, and use the other six keys for the
new stuff (sidestepping, zooming the camera).

Aug 17, 2003, 12:04pm

[View Quote] Not so true.

"trepanning" ... look it up.

http://www.braceface.com/medical/Pages/Trepanning.htm

Aug 17, 2003, 2:24am

Sure would round out the rest of the stuff that's in there now.

It would be lovely... tourist versions of all your precious models,
and versions only the real citizens can see.

Might even spur more people to sign up. Would surely make model
makers and world owners happier. At least people from other universes
would have to aqcuire a full on cit to steal things, and you could have a
list of allowed citizens on your OP anyway and only allow the specific ones
you want to access your path to see your models and stuff.

Just a thought. Noone will like the idea of course.

It can still be done anyway using other means. It would just be so
convenient if the cit number were there.

Aug 17, 2003, 3:53am

I'm not sure an SDK app would work. The referer string is being supplied by
the
browser and it could even just be making the whole referer string up
locally. There's
no real reason for it to be coming from the server.

An SDK app couldn't supply a different referer from what the browser is now
doing
on behalf of everyone who visits a given world.

A moddable server might do the trick, but only if it's the server that is
supplying
the referer string to the browser, and that is not certain.

If the browser did supply the cit, you could do cool things with just your
average
FTP only world setup, and without complex session management on the server.
It could just be an extra option in the new world options setup "Tack cit
number
on end of referer string".

[View Quote]

1  2  3  4  5  6  ...  13  |   Order by default