Yeah, so its hard to sell initially as anything but. IFF (if and only if) sony does a good job of making a media machine as well, then maybe we'll see some consumer saturation grow, create an emerging market. Otoh, a $400 media machine is very hard to justify. Unless you're comparing it to the extensible PC media center platform, in which case its a bargain, if limited to whatever shitty sony-centric drm-laden BS they turn it into. Frankly, I'll bite the bullet and point fingers now, Sony's going to have all the cards to do it right but the product isnt going to live up to expectations.
right now, i agree with you 100%, 10g is hardly needed anyhwere. but to address your orignal post, its that future thing that has me wondering. doubling up or tripling up on gigabit is currently standard practice for data intensive systems-- not just big iron, but rouge hackers cobbling together powerful clusters and grids for fun or profit. and it makes sense, its so damned cheap why not? besides, who can afford 10g? but with the throughput wars, how long can this really last? this is grassroots. this is mainstream, within the domain of computing hardware systems. for the actual computer industry, 1g is woefully insufficient.
grass-roots is people using drbd network replication with Xen to support live virtual host migration. if a filesystem fails, just migrate the hosts on that filesystem over to the host on the network backup and run them there. This sort of advanced system shuffling used to be the domain of blade systems and IBM, but now that bandwidth is becoming commodified and abundant we can start doing these things grassroots. Even now, some casual idiot can throw twenty one hard drives into a case for a couple terabytes of online storage. Soon with SAS (good overview), custom build storage will become only more of a reality.
Actually, SAS expanders use 4g infiniband interconnect. Maybe we just need cheaper infiniband. 4G is "nearly" enough.
So, currently storage solutions and blade systems are proprietary and expensive. With 10g and rapidly accelerating high availability and distributed systems, the linux kiddies are building it themselves.
this being said, i do wish to emphasize once more that I do really agree with you. there wasnt a single thing i didnt say yes yes and nod my head to in the parent.
I did some quick googling, and it seems Dashboard doesnt support SVG. Thats one of the biggest draws for me, being able to really build a complete UI "simply". Compiling a UI out of CSS, on the other hand, sounds "not fun". I wouldnt know where to begin piecing things together.
What methods of communication do dashboard widgets have? Standard Safari class XmlHttpRequest, I assume?
I'd be interested to see how much DOM support Dashboard has. Try some of the examples on Opera's widget devel; I'd be impressed if it supported mutable transparency.
supposedly the new ones also dont sound like jet engines.
the original 36 gig ones were strange beasts indeed. very quirky. sometimes they worked great though. great for doing fileserve. just make sure to reserve some bandwidth for yourself, eh?;-]
Is stackless still moving on its own? Looked kinda lame (as in sans legs) last time I gave it a poke.
The thing I dont understand is why stackless... I guess they just mean no C stack really. It seems like many programming languages (Boo for dotnet) can do continuations/generators/&all without going stackless. From the Continuations and Stackless Python doc I count find any compelling "stackless" features that really stuck out. We've had these things in C for a while now, they're called function pointers?
I too was wondering about how Eve was doing w/ stackless. I had a month straight were I flooded CAS (a Gallente noob corp) with chatter on the what Eve was running under & large scale architecting.
svg renderer, html & css engine, javascript engine, irc bitorrent & imap protcols, xml decoder, jpeg/png/gif decoders, caching algorithms, theme engine... pretty good really. most svg libs alone are over a meg, and they're just rasterizers, incapable of doing picking or animation. the raw.txt docs for these standards are probably a good 3 megs, even if you throw out the flavortexts.
the browsers you were talking about include lynx. thats about it.
actually wait, i used to use some fancy graphical bootloader. i think they did eventually include a web browser.:/ definately under a meg.
ya, i wget'd opera four times before i realized the download for TP2 really was only 4.0 MB.
shows in the runtime too. 224 mb linux laptop and i can still run 200 tabs fine. lightweight, well built, nicely async design. if for no other reason, i use opera because its the only thing that works.
On the other hand, your future mobile phone shouldnt have a problem. OpenGL ES is a mobile spec for OpenGL. Xegl is the Xgl for OGL:ES, iirc. Also, mesa is supposed to be able to do everything in software. Slowly.
OpenGL supposedly can be done over network. A number of really big really expensive display wall solutions do this. I think it'd be awsome to try and run remote X locally through an OpenGL network pipe display driver. I'm sure there would be *cough* problems *cough*, but it might be able to work. Not sure where to get info on GL over network...
But this is where it get really really interesitng: More interestingly is the possibilities crafted by indirect rendering. Since you're doing render to texture and then compositing, it'd be easy as shit to simply compress the textures in the video card via some shader, export the compressed results, then pipe them over the network to another target. [NoMachine's nx-compression basically does this, but it does it on the CPU, without indirect render.] Its not X, its not ogl, but it is pretty cool.
So basically you run the X app on the remote machine, and send the results out over the network. It'd be cool because you could switch where each individual app appears. Its just some indirect texture buffer after all. You could change the render target real time from remote to local, or display it on both or even three systems. And you could change each app individually. Hello [b]Rebuilding Xinerama[/b], without any of the nasty driver hackery. Xinerama as a window manager. If you spent the effort building it properly from day 1, it'd be really fairly easy to make it so the remote X server could be running on old technology.
Raptor SATA can break 88 MBps. Thats 704 Gbps. Add a little bit of network overhead, even for something as raw-metal as ATA-over-IP, you're still talking "damned close". Thats for _one_ hard drive, how passe.
Of course 10 gb is not for "consumers". At least not now, not for what we've got. But some day, if homes really are shuffling around 1080p streams (or bigger) like candy, even your average consumer home might need 10 gb backbones. If 1 gigabit is one single modern hard drive, 10 gigabit is not unreasonable for even a home backbone.
For enterprises over 100 employees doing data intensive work, I can picture 10 gigabit being a problem. They need someplace to go. In the meanwhile, the only way anyone else is going to be afford 10 gig is by there being some higher realm start the downpush of commodification.
---- Good call on the consumer gear. I am rather pissed the wifi routers are all still 100 megabit. I'm hoping with 802.11n we get a new round of routers with gigabit and cool new new processors. My WGT634U can only sustain 1.8 MBps from the USB to the wired ethernet. I doubt the USB to the wifi would be any better. Thats some shite DMA performance. On the other hand, dumb gigabit switches are pretty cheap, even some of the ones with Jumbo frames. But I'd still argue that gigabit has reached saturation, its cost is low enough that its basically just cause corps would rather use $0.30 chips rather than $1.30 chips in their routers and no one demands better for a do-it-all wirelessinternet widget.
Dual channel DDR2 at 1066 mhz can hit 6.4 Gbps. More and more systems are coming with a complete dual-channel hookup per cpu-core (a few with more), so for a four way K8 system thats twice the bandwidth of 100 gigabit ethernet. Of course, K8's dont run DDR2, when they do I doubt we'll see 1066 for a while, and DDR2 has such pathetic latencies (5-10-15-20 latencies anyone?) you'd be lucky to get five eights the theoretical. With four way though, thats still more than 100gigabit could manage.
With widgets (new in Opera9 Tech-Preview 2, examples here), opera is not only a web browser, its truly an application platform, capable of running independent programs. The current examples are true to the namesake, widgets, little gizmo's to show time &what not, but its the first time the graphical web is begin used as more than a browser page.
A little bit of polish is needed, but basically it's the only platform available for developing real SVG applications.
I hope Opera attempts to bring some real conformance to this entirely new class of web application. If it gets too proprietary its useless.
What is the differentce with the SR-80's? I wore my SR-80's well past the point of death, they were my alltime favorites. I vaguely understand what you're saying-- even at full volume some of my various playing devices were struggling (others were fine)-- but what is the actual difference that causes this?
So, to the asker, keep in mind that the open stage nature of the Grado's is double edged. You can hear whats going on (I'd attend lectures with them on back in college), but other people around you sometimes hear what you're listening to as well sometimes. If you're in a relatively silent office, it might stick out.
I once got sent to the office for discussing capacitors with a friend way back in high school. We were looking at building a pretty big Leech amp, and the guy writing the guide we were looking at kept referring to how mnay "dog feet" the capacitor had; how far the potential energy of the cap could hypothetically throw his black lab.
We found some 10 farad 60V caps on Fair Radio Supply. Never bought any, too cheap to work. Course, my college roomate was best friends with the campus supply depot, so he had some eighty pound 600V and 1kv caps that he wanted to turn into a coilgun someday.
The Casio lets you save settings. When you save settings, it takes a thumbnail of the first shot you take with that settings. Browing across settings is just a posterboard of different shots, nine to a screen, so you know what conditions you were aiming for when you saved the setting. Its amazingly efficient at finding good previously-used-settings because you see what you were trying to shoot then. It also comes iwth a dozen pretty good presets with very good baseline settings.
When it comes down to it, this is really one of the key differences between the prosumer and consumer... whether you're just snapping in auto or whether you're doing any tuning. A good small prosumer can shoot amazing shots, but most people just use em point and click.
The too small is a point. Sometimes I think it'd be cool to see gyroscopically stabilized cameras, but making small hihg inertia gyros would be... a "fun" engineering project. Of course, you can just crank your shot to over 1/60 or 1/100 and you should be ok, but that impacts the low light capabilities a lot.
All real cameras support RAW, so its fairly moot, but the idea that you would throw away the RAW data is scary. Usually the algorithms deduce what's going on fine, but for the pathological cases, RAW is invaluable. Sometimes the Bayes reconstruction is just plain wrong for a tricky image, leading to a pretty unnatural looking image.
I really enjoy your points, you're dead on. But, just for the record here, I dont like the first ISO3200 shots you linked at all, you can see the noise all over her arm (disregarding the shake). I suspect in part this is one of those pathological cases where the algorithms' working on the Bayesian data just got it wrong, but some of it is likely thermal too. Otoh, the indoor shot at the end does look really smooth, and I'd be interested on a fullsize of the ISO1600 water-theme, although I suspect they're flawless.
embedded systems need time to evolve first. right now a solution would be full custom, pricing it very high. we need some commoditazation. i'd place that around 2007, but the bandwagons going to move fast once it gets here.
look around in three years; computers will have undergone a throughput revolution. pci-expres will achieve penetration into the embedded market so there'll be a low-trace-count high-bandwidth low-voltage interconnect. the major semiconductor people just released the first round of reasonable priced off the shelf H.264 encoding (ti's da vinci for example). Current solutions needed either high bandwidth raw storage or proprietary encoders, all of which are in the 'custom' domain.
I dont know where CCD's and the op-amps stand. I'd wager they're a bit of a limiting factor. I'd be the thermal issues become problematic when you're running a CCD at over 20 Mbps of raw data. Active cooling is power intensive.
Bayer pattern is only useful if your camera outputs RAW formats. Otherwise you have to trust all in a couple firmware engineers. So you either have to go looking, or buy a $500+ camera to get RAW output. Thats my qualm with your Bayer defense, although you are basically right. Interpolation is a wonderful thing. Unless you're shooting pathological cases. They happen. Without manual tweaking you're basically fucked.
Your flat statement
A 6mp Bayer sensor in an DSLR is already better than 35mm film. By 10MP it is significantly better.
is utter rubbish, completely nonsensical. 35mm what? Film has grain & grain determines its resolution. What ISO? whose film? Are you going to tell me a 10mp is better than some 25 ISO ultra-fine B&W? Is 3200 iso 35mm worse than a 10mp for lowlight indoors? 35mm is amazing because its adaptable.
Slashdot Mirror
#1 is accountability. know what's actually happening.
most people have no idea how long they spend on their projects. it makes it very hard to appraise anything.
Yeah, so its hard to sell initially as anything but. IFF (if and only if) sony does a good job of making a media machine as well, then maybe we'll see some consumer saturation grow, create an emerging market. Otoh, a $400 media machine is very hard to justify. Unless you're comparing it to the extensible PC media center platform, in which case its a bargain, if limited to whatever shitty sony-centric drm-laden BS they turn it into. Frankly, I'll bite the bullet and point fingers now, Sony's going to have all the cards to do it right but the product isnt going to live up to expectations.
Myren
Yeah, they'd obviously done a great job of protecting us so far!
right now, i agree with you 100%, 10g is hardly needed anyhwere. but to address your orignal post, its that future thing that has me wondering. doubling up or tripling up on gigabit is currently standard practice for data intensive systems-- not just big iron, but rouge hackers cobbling together powerful clusters and grids for fun or profit. and it makes sense, its so damned cheap why not? besides, who can afford 10g? but with the throughput wars, how long can this really last? this is grassroots. this is mainstream, within the domain of computing hardware systems. for the actual computer industry, 1g is woefully insufficient.
grass-roots is people using drbd network replication with Xen to support live virtual host migration. if a filesystem fails, just migrate the hosts on that filesystem over to the host on the network backup and run them there. This sort of advanced system shuffling used to be the domain of blade systems and IBM, but now that bandwidth is becoming commodified and abundant we can start doing these things grassroots. Even now, some casual idiot can throw twenty one hard drives into a case for a couple terabytes of online storage. Soon with SAS (good overview), custom build storage will become only more of a reality.
Actually, SAS expanders use 4g infiniband interconnect. Maybe we just need cheaper infiniband. 4G is "nearly" enough.
So, currently storage solutions and blade systems are proprietary and expensive. With 10g and rapidly accelerating high availability and distributed systems, the linux kiddies are building it themselves.
this being said, i do wish to emphasize once more that I do really agree with you. there wasnt a single thing i didnt say yes yes and nod my head to in the parent.
his wings are feeling better this time?
he was planning on setting a new record a couple months ago with global flyer, but the crew scrapped up one of he wings, iirc.
3530 lbs unloaded. Pretty cool.
Interesting. I honestly had no idea.
I did some quick googling, and it seems Dashboard doesnt support SVG. Thats one of the biggest draws for me, being able to really build a complete UI "simply". Compiling a UI out of CSS, on the other hand, sounds "not fun". I wouldnt know where to begin piecing things together.
What methods of communication do dashboard widgets have? Standard Safari class XmlHttpRequest, I assume?
I'd be interested to see how much DOM support Dashboard has. Try some of the examples on Opera's widget devel; I'd be impressed if it supported mutable transparency.
Thanks tho, I really had no idea about dashboard.
supposedly the new ones also dont sound like jet engines.
;-]
the original 36 gig ones were strange beasts indeed. very quirky. sometimes they worked great though. great for doing fileserve. just make sure to reserve some bandwidth for yourself, eh?
Is stackless still moving on its own? Looked kinda lame (as in sans legs) last time I gave it a poke.
The thing I dont understand is why stackless... I guess they just mean no C stack really. It seems like many programming languages (Boo for dotnet) can do continuations/generators/&all without going stackless. From the Continuations and Stackless Python doc I count find any compelling "stackless" features that really stuck out. We've had these things in C for a while now, they're called function pointers?
I too was wondering about how Eve was doing w/ stackless. I had a month straight were I flooded CAS (a Gallente noob corp) with chatter on the what Eve was running under & large scale architecting.
Myren
svg renderer, html & css engine, javascript engine, irc bitorrent & imap protcols, xml decoder, jpeg/png/gif decoders, caching algorithms, theme engine... pretty good really. most svg libs alone are over a meg, and they're just rasterizers, incapable of doing picking or animation. the raw .txt docs for these standards are probably a good 3 megs, even if you throw out the flavortexts.
:/ definately under a meg.
the browsers you were talking about include lynx. thats about it.
actually wait, i used to use some fancy graphical bootloader. i think they did eventually include a web browser.
ya, i wget'd opera four times before i realized the download for TP2 really was only 4.0 MB.
shows in the runtime too. 224 mb linux laptop and i can still run 200 tabs fine. lightweight, well built, nicely async design. if for no other reason, i use opera because its the only thing that works.
On the other hand, your future mobile phone shouldnt have a problem.
OpenGL ES is a mobile spec for OpenGL. Xegl is the Xgl for OGL:ES, iirc.
Also, mesa is supposed to be able to do everything in software. Slowly.
OpenGL supposedly can be done over network. A number of really big really expensive display wall solutions do this. I think it'd be awsome to try and run remote X locally through an OpenGL network pipe display driver. I'm sure there would be *cough* problems *cough*, but it might be able to work. Not sure where to get info on GL over network...
But this is where it get really really interesitng:
More interestingly is the possibilities crafted by indirect rendering. Since you're doing render to texture and then compositing, it'd be easy as shit to simply compress the textures in the video card via some shader, export the compressed results, then pipe them over the network to another target. [NoMachine's nx-compression basically does this, but it does it on the CPU, without indirect render.] Its not X, its not ogl, but it is pretty cool.
So basically you run the X app on the remote machine, and send the results out over the network. It'd be cool because you could switch where each individual app appears. Its just some indirect texture buffer after all. You could change the render target real time from remote to local, or display it on both or even three systems. And you could change each app individually. Hello [b]Rebuilding Xinerama[/b], without any of the nasty driver hackery. Xinerama as a window manager. If you spent the effort building it properly from day 1, it'd be really fairly easy to make it so the remote X server could be running on old technology.
Indirect rendering fucking rocks.
Myren
Doesnt it use indirect rendering (render to texture)? Is that available on old Geforce1 class or TnT hardware?
Dont worry, the de-evolutions are nearly perfectly analogous.
I'd be interested to see comparison of r300 feature support v. nvidia & ati's drivers.
Raptor SATA can break 88 MBps. Thats 704 Gbps. Add a little bit of network overhead, even for something as raw-metal as ATA-over-IP, you're still talking "damned close". Thats for _one_ hard drive, how passe.
Of course 10 gb is not for "consumers". At least not now, not for what we've got. But some day, if homes really are shuffling around 1080p streams (or bigger) like candy, even your average consumer home might need 10 gb backbones. If 1 gigabit is one single modern hard drive, 10 gigabit is not unreasonable for even a home backbone.
For enterprises over 100 employees doing data intensive work, I can picture 10 gigabit being a problem. They need someplace to go. In the meanwhile, the only way anyone else is going to be afford 10 gig is by there being some higher realm start the downpush of commodification.
----
Good call on the consumer gear. I am rather pissed the wifi routers are all still 100 megabit. I'm hoping with 802.11n we get a new round of routers with gigabit and cool new new processors. My WGT634U can only sustain 1.8 MBps from the USB to the wired ethernet. I doubt the USB to the wifi would be any better. Thats some shite DMA performance. On the other hand, dumb gigabit switches are pretty cheap, even some of the ones with Jumbo frames. But I'd still argue that gigabit has reached saturation, its cost is low enough that its basically just cause corps would rather use $0.30 chips rather than $1.30 chips in their routers and no one demands better for a do-it-all wirelessinternet widget.
You'd need entire google datacenters to layer7 packet shape a 100 gigabit stream.
Here's to hoping our bandwidth exceeds moore's ability to packet shape it!
100 / 8 = 12.5
Dual channel DDR2 at 1066 mhz can hit 6.4 Gbps. More and more systems are coming with a complete dual-channel hookup per cpu-core (a few with more), so for a four way K8 system thats twice the bandwidth of 100 gigabit ethernet. Of course, K8's dont run DDR2, when they do I doubt we'll see 1066 for a while, and DDR2 has such pathetic latencies (5-10-15-20 latencies anyone?) you'd be lucky to get five eights the theoretical. With four way though, thats still more than 100gigabit could manage.
Hello 2006? Let the throughput wars begin!
Myren
With widgets (new in Opera9 Tech-Preview 2, examples here), opera is not only a web browser, its truly an application platform, capable of running independent programs. The current examples are true to the namesake, widgets, little gizmo's to show time &what not, but its the first time the graphical web is begin used as more than a browser page.
A little bit of polish is needed, but basically it's the only platform available for developing real SVG applications.
I hope Opera attempts to bring some real conformance to this entirely new class of web application. If it gets too proprietary its useless.
Myren
What is the differentce with the SR-80's? I wore my SR-80's well past the point of death, they were my alltime favorites. I vaguely understand what you're saying-- even at full volume some of my various playing devices were struggling (others were fine)-- but what is the actual difference that causes this?
So, to the asker, keep in mind that the open stage nature of the Grado's is double edged. You can hear whats going on (I'd attend lectures with them on back in college), but other people around you sometimes hear what you're listening to as well sometimes. If you're in a relatively silent office, it might stick out.
I once got sent to the office for discussing capacitors with a friend way back in high school. We were looking at building a pretty big Leech amp, and the guy writing the guide we were looking at kept referring to how mnay "dog feet" the capacitor had; how far the potential energy of the cap could hypothetically throw his black lab.
We found some 10 farad 60V caps on Fair Radio Supply. Never bought any, too cheap to work. Course, my college roomate was best friends with the campus supply depot, so he had some eighty pound 600V and 1kv caps that he wanted to turn into a coilgun someday.
The Casio lets you save settings. When you save settings, it takes a thumbnail of the first shot you take with that settings. Browing across settings is just a posterboard of different shots, nine to a screen, so you know what conditions you were aiming for when you saved the setting. Its amazingly efficient at finding good previously-used-settings because you see what you were trying to shoot then. It also comes iwth a dozen pretty good presets with very good baseline settings.
When it comes down to it, this is really one of the key differences between the prosumer and consumer... whether you're just snapping in auto or whether you're doing any tuning. A good small prosumer can shoot amazing shots, but most people just use em point and click.
The too small is a point. Sometimes I think it'd be cool to see gyroscopically stabilized cameras, but making small hihg inertia gyros would be... a "fun" engineering project. Of course, you can just crank your shot to over 1/60 or 1/100 and you should be ok, but that impacts the low light capabilities a lot.
All real cameras support RAW, so its fairly moot, but the idea that you would throw away the RAW data is scary. Usually the algorithms deduce what's going on fine, but for the pathological cases, RAW is invaluable. Sometimes the Bayes reconstruction is just plain wrong for a tricky image, leading to a pretty unnatural looking image.
I really enjoy your points, you're dead on. But, just for the record here, I dont like the first ISO3200 shots you linked at all, you can see the noise all over her arm (disregarding the shake). I suspect in part this is one of those pathological cases where the algorithms' working on the Bayesian data just got it wrong, but some of it is likely thermal too. Otoh, the indoor shot at the end does look really smooth, and I'd be interested on a fullsize of the ISO1600 water-theme, although I suspect they're flawless.
I believe this rote is called Virtual Holography.
embedded systems need time to evolve first. right now a solution would be full custom, pricing it very high. we need some commoditazation. i'd place that around 2007, but the bandwagons going to move fast once it gets here.
look around in three years; computers will have undergone a throughput revolution. pci-expres will achieve penetration into the embedded market so there'll be a low-trace-count high-bandwidth low-voltage interconnect. the major semiconductor people just released the first round of reasonable priced off the shelf H.264 encoding (ti's da vinci for example). Current solutions needed either high bandwidth raw storage or proprietary encoders, all of which are in the 'custom' domain.
I dont know where CCD's and the op-amps stand. I'd wager they're a bit of a limiting factor. I'd be the thermal issues become problematic when you're running a CCD at over 20 Mbps of raw data. Active cooling is power intensive.
The tech's getting here...
Your flat statement
is utter rubbish, completely nonsensical. 35mm what? Film has grain & grain determines its resolution. What ISO? whose film? Are you going to tell me a 10mp is better than some 25 ISO ultra-fine B&W? Is 3200 iso 35mm worse than a 10mp for lowlight indoors? 35mm is amazing because its adaptable.