Despite the troubles of RLX and related companies (probably due to the general market downturn more than their specific product), the server market for low power chips will come back.
At some point we'll probably see benchmarks on TPC/kW or Webstones/rack where Transmeta could make a dent.
An intriguing question. What I wonder is, if server power becomes such a significant aspect of design due to the price of electricity, what does that mean for the clients of these systems? I wonder if it implies that the clients will be using those systems less themselves due to constraints on their own energy usage. In other words, server power considerations and client power considerations (cost, availability, reliability) may be on a parallel track, with client demand inversely proportionate to client power considerations.
When power is perceived as 'free' as it is in the US (at least relative to PCs), then nobody cares and only the most intensely dense and large data centers make it much of a calculation. When power is 'expensive', clients will make less use of their PCs, demanding less from the servers, and automatically reducing data center power demands.
I realize this is a little over simplistic, but its kind of a macroeconomic way of looking at it. There are also probably specialized environments were power and heat may matter a lot (third world, remote areas, ships/planes, etc). And it may be that future computing models shift computational intensity to the network and rely more on thin clients (as pundits have been saying for years), and this may actually be *more* true as thin, mobile clients increase capabilities.
Even in power-scarce situations, users may switch to thin clients, partially in an attempt to shift their power burden to someone else.
I work in advertising and I've asked creative directors if they take "zipping" (the term most used for FFing commercials around here) into account when planning an ad; more long shots of product names or usage, and fewer jump cuts and scene changes so that the message remains coherent at FF speeds.
They've all said "NO", but a couple indicated they thought it would be worth investigating. But since the spots are all evaluated by the client at 1x, it's hard to make arguments about 15x comprehension when you can't sell basic items at 1x..
now await all the flames which will call me an idiot for not understanding Exchange server.
I don't think you're an idiot, but its the glue code between LDAP, IMAP, the MTA that makes most of the magic happen. Plus you've left out calendaring, which I personally would think gets split between IMAP and the MTA, but the MTA/IMAP model doesn't suggest a natural place for calendaring, which is a huge reason these kinds of systems get deployed.
Lots of managers (including two CIOs I know) argue that management is a discpline unto itself and knowing how to manage is enough; technical knowledge enough to make decisions is learned or imparted through employees or consultants.
I buy this to a certain extent, but I've also seen CIOs make horrible technology decisions because they didn't trust their own people, were misled by consultants, or just plain made decisions because they *thought* they knew. I'm pretty sure the latter is a big culprit, as are CIOs that drag in consultants who disagree with their staff and create a huge we/they problem. And then there are bottom-liners who manage to the bottom line, "trimming costs" as a sign of "good management."
I also think that an organization has to be structured in such a way that good, consultive management can work. Frequently it's not structured that way, management doesn't trust employees, employees don't trust management, and the whole process of decisionmaking gets flushed down the toilet.
I personally think that effective management requires a lot of experience in the field you're managing AND a solid management training background. Past "Experts" now in management can certainly micromanage or get into situations where they override their technical people simply because "in their day" things were done differently.
I guess my whole point was that even if 128k MP3s are horrible (I seriously doubt that the majority of people would consider them sounding bad in a standalone environment, but that's another argument), the *listening environment* has a pretty high noise floor of its own which both psychologically distracts the listener and prevents you from being able to make critical aural distinctions.
I'd presume that people who have taught themselves to listen very, very closely to 128k MP3s in quiet environments on good equipment may be able to make occasional distinctions. I don't think though that ordinary people in ordinary listening situations would make those distinctions, though.
Most of my listening is done two places, either walking with a portable or driving in my car. My portable is a Teac mini-CD MP3 player (plays ISOs with MP3s on the small CDs), my car is the CD player that came with it ('01 Honda CR-V).
While walking outside, there's so much background noise and my headphones (either the earbuds or my Koss Porta-Pro juniors) just aren't good enough to be able to tell a "really good" MP3 from an OK one. I have a ton encoded with the Xing encoder (most were done with VBR), and I couldn't tell you which ones were which, they sound fine to me. I can tell *damaged* MP3s (those with skips, bad warble) but not "bad" ones.
It's even more true in my car. My car CDs are audio CDs but made from MP3 files. The car is such a noisy environment, that I don't see how you could tell a "good" MP3 from a "bad" one (again, damaged is another category).
I think you'd have to do some serious, high-quality headphone listening in a really quiet environment to be able to tell the difference. I think the vast majority of people might be able to tell the difference in some places with A/B listening when coached, but if you just put the MP3 on and played it they'd never say "that sounds off, is it a 128k xing?"
Please name one Republican-supported birth control initiative that doesn't start and end with "abstinance". I'd agree with the sentiment that minors shouldn't get access to birth control without parental consent, but abstinence only "birth control education" is about as rational as "creationism-only biology."
IMHO the anti-abortion movement is primarily an 'anti-sexuality' movement who really won't be satisfied until they have made sexuality what it was in the 19th century.
The LDS church has *recently* taken a more public stand against polygamy, but the fact that its so widespread (a recent NYT article suggested 10s of thousands in Utah are in polygamous marriages) indicates that church "opposition" to it has been muted for a long time. The recent Elizabeth Smart case will focus some rather unpleasant publicity on LDS practices and attitudes, even if they aren't "official."
Besides, as a parent poster noted, they had an anti-porn czar. If polygamy is so bad, why wasn't there an anti-polygamy czar? Sounds to me like it was more ignored or looked over than actively banned.
Not an option at all. I cannot get outside at the 45 story highrise downtown building our offices are in. We are tenants of about 6 floors, there is no "outside" that belongs to us or we have access too.
The same is true in the other sites I manage; tenants of large, otherwise sealed buildings.
I'd love to do what you suggest, but it is simply not an option. I'd be just thrilled with the nearest airport temperature data.
Is there a command line tool for grabbing weather data? I'd like to do some MRTG graphs of internal temp sensors vs. external temperatures. Unfortunately I don't have a way to get "outside" at the facilities I want outside temps for. If there was some "wxfetch --temp --site=sna" command that would give me easily parsable basic weather data (no pix, all text) it'd rock.
I like the twisty logic many of the LDS do. It's not ok to watch porn, but it is OK to satisfy your sexual needs by "marrying" multiple teenage girls. Nice.
And then there's all the hotties that troll temple square and try to sway new converts. I had fun with them, though -- "I like sex a lot. Is that a problem in your religion?" I get the impression that its not -- the women are taught to satisfy the men any way they want. I'll bet there's more cornholing in LDS marriages than others...
First Ammendment: Dems want to limit this so you can say whatever you want, as long as you never say anything that might insult even the most ignorant person of "minority standing." Can't talk about black racism, homosexual promiscuity or anything else that might offend them. Give the republicans back half a point on this one, but the overall freedom of speech score goes to the dems.
Abortion: Its worse than you say. Republicans are against birth control, which is the best way to prevent abortion. They're also against childcare and other things you need to raise that child they want you to have.
Labor: Dems lose points here. They talk pro-worker, but at the same time their multiculturalism and pro-immigrant stance prevents them from taking a line on massive immigration which undermines native-born workers. Republicans are notoriously quiet on this one as well.
Jail vs. Education: They both lose. Since we can't punish kids in school, nobody gets a good education. We can't punish them due to democratic multiculturalism -- kicking out the rowdy kids would kick out mostly blacks. We need both here -- kick the disruptive kids out of school. If it leads to jail, tough shit. Rich private schools are only of real benefit because their kids will shut the fuck up, pay attention and let everyone else learn.
Environment: I hate to see the land raped, but at the same time we gotta have houses. The left would have us crammed into massive Stalinist concrete apartment complexes instead of building single family homes ("High Density Urbanism"). And the left is generally anti development of any kind. It'd be nice to have a compromise of some kind that wasn't either side's extreme.
I do agree with your overall conclusions. The rich get richer, everyone else gets cornholed.
Re:It matters that Microsoft bought it.
on
Virtual PC 6 Review
·
· Score: 2, Interesting
I wouldn't be surprised if future Mac products were just the windows versions integrated with an application-specific VirtualPC wrapper.
This way they don't even have to make Mac-specific applications at all, just maintain the wrapper.
I don't know how performance-inducing this would be, but it'd save a ton of development time.
I'd like to see a flavor of this for FPS games. I'd like to be able to play as a single actor in a huge group battle (civil war, roman era, whatever) and have good ai opponents and teammates.
I've often wondered if it was a rendering hurdle or an AI processing problem.
You get access to the data that the 'normal' machine has access too and can presumably 'borrow' its IP address and other logical connection attributes, but I don't see how it lets you bypass any security that would otherwise apply to any other device.
The device you describe would be great for spying on a specific PC, especially if it could fit into a wall socket. Add a cell modem and you have some seriously compromised security on that PC.
Sniffing data assumes they are able to sniff data in the first place. Using good switches that don't turn into dumb hubs when their packet buffers or MAC tables flood is a good idea.
A properly and sanely layer-3 segmented network helps here, too -- keep your servers on a more secure core network and require traffic going from one subnet to another to traverse the core is another good idea, as you can limit what can happen between subnets.
Don't allow incoming connections to client subnets ever, and seriously limit non-proxied outbound connections.
I haven't done this, but I'd also be amused with the idea of DHCP leases that require an IP change periodically (ala the cable modem people), or at least check your DHCP leases to find "old" leases that have't turned over.
It may be possible to reduce a "sleeper" bug like this to unusability from the outside. Although if they embed a cellular modem in it, all bets are off.
IIRC they test newborns to see if they are "PK babies" right away, since they'll die before they ever get their first diet coke due to phenylalanine exposure in regular foods.
My friends from the fringe are constantly chiding my use of Nutrasweet, but honestly I like it better than sugar and it doesn't make me feel all weird like lots of sugar does. The only other sweeteners I use besides nutrasweet are honey and molases, but these are flavorizers as much as sweeteners.
I just wish nutrasweet was more shelf-stable and high temperature stable so I could substitute it for sugar in recipes and not have the diet coke in the back of the fridge go south.
Once every few years? I upgrade more than that, at least to keep up with the bind bugs and sendmail bugs. Yes, I could binary update those specific bits with patches, but there's nothing to be lost with a buildworld in my experience (4.0 through 4.8-RC), and it keeps everything coherent. It seems like keeping up is easier if you at least follow the minor point updates to stable as well, as occasionally something interesting gets MFC'd requiring a little more work.
The buildworld process is about as unsupervised as a Win2k service pack. There's a few build commands, a mergemaster run, and then a reboot. It doesn't require total attention, but its not 100% background either. And once in a great while it breaks, requiring some investigating to see what's broken (me or world).
Perhaps braver people than I just do binary updates for bug fixes and only rebuild world every few years, but I like it more frequently than that. We usually do enough hardware updates that there's some motivation at least every 18 months to swap around boxes, which usually involves at least some kind build if not a total reinstall.
My experience managing Win2k (12 servers, 3 locations) is that it's really quite simple to manage and even simpler to keep updated. It helps to have high speed (256k, minimum) connectivity to the sites for terminal services (reboots, service packs, etc).
A buildworld on one of my FreeBSD boxes takes much longer than even a Win2K SP with backup, plus there's time spent re-building stuff from ports when that needs updating.
I'd say they're roughly equal in terms of time; FreeBSD needs updating less often, but takes longer and is more involved. Win2k is quicker to update, but you have to do it more often.
If *ALL* I did was manage the boxes (OS updates, hardware maintenance, upgrades), I could do an equal number of Win2k and FreeBSD machines, and probably 2-3 times what I do now. But I also have to do all network infrastructure (routers, switches, dialup, cabling) as well as deal with the politics, PHBs and end users, all of which sucks valuable time.
Multiple desktop 3D GUI
on
Opencroquet
·
· Score: 1
I'm not a huge fan of traditional multidesktop environments because they're too geographically distinct -- I have to "switch" between them, and that involves an annoying redraw penalty as well as a mental "reset" to refamiliarize myself with that environment.
I'm not sure get the 3D UI completely, but what would make sense to me would be putting the user inside of a many-sided polygon container. Each interior face of the sphere would be a typical desktop. The controls for the GUI would enable zooming out to see multiple faces/desktops simultaneously, as well as the ability to pan over desktops, in addition to being able to "lock" onto a desktop. Windows could be moved between desktops all-at-once, or dragged if you were panned over them.
It would have interesting multimonitor potential, as well -- the monitors could be assigned as geographically adjacent so that they showed adjacent polygonal surfaces/desktops, the polygonal surface could be sized to the combined monitor resolution or the monitors could be 'detatched' so that they showed different views (perhaps one zoomed out, one locked).
I don't think this would be that hard to implement, either (disclaimer: not a developer), since it wouldn't involve changes in the base GUI, just in the way that desktops as a whole are presented and navigated.
Slashdot Mirror
Despite the troubles of RLX and related companies (probably due to the general market downturn more than their specific product), the server market for low power chips will come back. At some point we'll probably see benchmarks on TPC/kW or Webstones/rack where Transmeta could make a dent.
An intriguing question. What I wonder is, if server power becomes such a significant aspect of design due to the price of electricity, what does that mean for the clients of these systems? I wonder if it implies that the clients will be using those systems less themselves due to constraints on their own energy usage. In other words, server power considerations and client power considerations (cost, availability, reliability) may be on a parallel track, with client demand inversely proportionate to client power considerations.
When power is perceived as 'free' as it is in the US (at least relative to PCs), then nobody cares and only the most intensely dense and large data centers make it much of a calculation. When power is 'expensive', clients will make less use of their PCs, demanding less from the servers, and automatically reducing data center power demands.
I realize this is a little over simplistic, but its kind of a macroeconomic way of looking at it. There are also probably specialized environments were power and heat may matter a lot (third world, remote areas, ships/planes, etc). And it may be that future computing models shift computational intensity to the network and rely more on thin clients (as pundits have been saying for years), and this may actually be *more* true as thin, mobile clients increase capabilities.
Even in power-scarce situations, users may switch to thin clients, partially in an attempt to shift their power burden to someone else.
I work in advertising and I've asked creative directors if they take "zipping" (the term most used for FFing commercials around here) into account when planning an ad; more long shots of product names or usage, and fewer jump cuts and scene changes so that the message remains coherent at FF speeds.
They've all said "NO", but a couple indicated they thought it would be worth investigating. But since the spots are all evaluated by the client at 1x, it's hard to make arguments about 15x comprehension when you can't sell basic items at 1x..
now await all the flames which will call me an idiot for not understanding Exchange server.
I don't think you're an idiot, but its the glue code between LDAP, IMAP, the MTA that makes most of the magic happen. Plus you've left out calendaring, which I personally would think gets split between IMAP and the MTA, but the MTA/IMAP model doesn't suggest a natural place for calendaring, which is a huge reason these kinds of systems get deployed.
Lots of managers (including two CIOs I know) argue that management is a discpline unto itself and knowing how to manage is enough; technical knowledge enough to make decisions is learned or imparted through employees or consultants.
I buy this to a certain extent, but I've also seen CIOs make horrible technology decisions because they didn't trust their own people, were misled by consultants, or just plain made decisions because they *thought* they knew. I'm pretty sure the latter is a big culprit, as are CIOs that drag in consultants who disagree with their staff and create a huge we/they problem. And then there are bottom-liners who manage to the bottom line, "trimming costs" as a sign of "good management."
I also think that an organization has to be structured in such a way that good, consultive management can work. Frequently it's not structured that way, management doesn't trust employees, employees don't trust management, and the whole process of decisionmaking gets flushed down the toilet.
I personally think that effective management requires a lot of experience in the field you're managing AND a solid management training background. Past "Experts" now in management can certainly micromanage or get into situations where they override their technical people simply because "in their day" things were done differently.
Much worse than the AT&T version. The words are run togther too much.
Haven't gotten the IBM one to work yet.
I guess my whole point was that even if 128k MP3s are horrible (I seriously doubt that the majority of people would consider them sounding bad in a standalone environment, but that's another argument), the *listening environment* has a pretty high noise floor of its own which both psychologically distracts the listener and prevents you from being able to make critical aural distinctions.
I'd presume that people who have taught themselves to listen very, very closely to 128k MP3s in quiet environments on good equipment may be able to make occasional distinctions. I don't think though that ordinary people in ordinary listening situations would make those distinctions, though.
Most of my listening is done two places, either walking with a portable or driving in my car. My portable is a Teac mini-CD MP3 player (plays ISOs with MP3s on the small CDs), my car is the CD player that came with it ('01 Honda CR-V).
While walking outside, there's so much background noise and my headphones (either the earbuds or my Koss Porta-Pro juniors) just aren't good enough to be able to tell a "really good" MP3 from an OK one. I have a ton encoded with the Xing encoder (most were done with VBR), and I couldn't tell you which ones were which, they sound fine to me. I can tell *damaged* MP3s (those with skips, bad warble) but not "bad" ones.
It's even more true in my car. My car CDs are audio CDs but made from MP3 files. The car is such a noisy environment, that I don't see how you could tell a "good" MP3 from a "bad" one (again, damaged is another category).
I think you'd have to do some serious, high-quality headphone listening in a really quiet environment to be able to tell the difference. I think the vast majority of people might be able to tell the difference in some places with A/B listening when coached, but if you just put the MP3 on and played it they'd never say "that sounds off, is it a 128k xing?"
Please name one Republican-supported birth control initiative that doesn't start and end with "abstinance". I'd agree with the sentiment that minors shouldn't get access to birth control without parental consent, but abstinence only "birth control education" is about as rational as "creationism-only biology."
IMHO the anti-abortion movement is primarily an 'anti-sexuality' movement who really won't be satisfied until they have made sexuality what it was in the 19th century.
An appealing idea, but not exactly simple on a glass curtainwall structure.
Besides, the washers at our building aren't just blue collar, I think they're fresh out of prison or something.
The LDS church has *recently* taken a more public stand against polygamy, but the fact that its so widespread (a recent NYT article suggested 10s of thousands in Utah are in polygamous marriages) indicates that church "opposition" to it has been muted for a long time. The recent Elizabeth Smart case will focus some rather unpleasant publicity on LDS practices and attitudes, even if they aren't "official."
Besides, as a parent poster noted, they had an anti-porn czar. If polygamy is so bad, why wasn't there an anti-polygamy czar? Sounds to me like it was more ignored or looked over than actively banned.
Not an option at all. I cannot get outside at the 45 story highrise downtown building our offices are in. We are tenants of about 6 floors, there is no "outside" that belongs to us or we have access too.
The same is true in the other sites I manage; tenants of large, otherwise sealed buildings.
I'd love to do what you suggest, but it is simply not an option. I'd be just thrilled with the nearest airport temperature data.
Is there a command line tool for grabbing weather data? I'd like to do some MRTG graphs of internal temp sensors vs. external temperatures. Unfortunately I don't have a way to get "outside" at the facilities I want outside temps for. If there was some "wxfetch --temp --site=sna" command that would give me easily parsable basic weather data (no pix, all text) it'd rock.
I like the twisty logic many of the LDS do. It's not ok to watch porn, but it is OK to satisfy your sexual needs by "marrying" multiple teenage girls. Nice.
And then there's all the hotties that troll temple square and try to sway new converts. I had fun with them, though -- "I like sex a lot. Is that a problem in your religion?" I get the impression that its not -- the women are taught to satisfy the men any way they want. I'll bet there's more cornholing in LDS marriages than others...
First Ammendment: Dems want to limit this so you can say whatever you want, as long as you never say anything that might insult even the most ignorant person of "minority standing." Can't talk about black racism, homosexual promiscuity or anything else that might offend them. Give the republicans back half a point on this one, but the overall freedom of speech score goes to the dems.
Abortion: Its worse than you say. Republicans are against birth control, which is the best way to prevent abortion. They're also against childcare and other things you need to raise that child they want you to have.
Labor: Dems lose points here. They talk pro-worker, but at the same time their multiculturalism and pro-immigrant stance prevents them from taking a line on massive immigration which undermines native-born workers. Republicans are notoriously quiet on this one as well.
Jail vs. Education: They both lose. Since we can't punish kids in school, nobody gets a good education. We can't punish them due to democratic multiculturalism -- kicking out the rowdy kids would kick out mostly blacks. We need both here -- kick the disruptive kids out of school. If it leads to jail, tough shit. Rich private schools are only of real benefit because their kids will shut the fuck up, pay attention and let everyone else learn.
Environment: I hate to see the land raped, but at the same time we gotta have houses. The left would have us crammed into massive Stalinist concrete apartment complexes instead of building single family homes ("High Density Urbanism"). And the left is generally anti development of any kind. It'd be nice to have a compromise of some kind that wasn't either side's extreme.
I do agree with your overall conclusions. The rich get richer, everyone else gets cornholed.
I wouldn't be surprised if future Mac products were just the windows versions integrated with an application-specific VirtualPC wrapper.
This way they don't even have to make Mac-specific applications at all, just maintain the wrapper.
I don't know how performance-inducing this would be, but it'd save a ton of development time.
I'd like to see a flavor of this for FPS games. I'd like to be able to play as a single actor in a huge group battle (civil war, roman era, whatever) and have good ai opponents and teammates.
I've often wondered if it was a rendering hurdle or an AI processing problem.
You get access to the data that the 'normal' machine has access too and can presumably 'borrow' its IP address and other logical connection attributes, but I don't see how it lets you bypass any security that would otherwise apply to any other device.
The device you describe would be great for spying on a specific PC, especially if it could fit into a wall socket. Add a cell modem and you have some seriously compromised security on that PC.
Sniffing data assumes they are able to sniff data in the first place. Using good switches that don't turn into dumb hubs when their packet buffers or MAC tables flood is a good idea.
A properly and sanely layer-3 segmented network helps here, too -- keep your servers on a more secure core network and require traffic going from one subnet to another to traverse the core is another good idea, as you can limit what can happen between subnets.
Don't allow incoming connections to client subnets ever, and seriously limit non-proxied outbound connections.
I haven't done this, but I'd also be amused with the idea of DHCP leases that require an IP change periodically (ala the cable modem people), or at least check your DHCP leases to find "old" leases that have't turned over.
It may be possible to reduce a "sleeper" bug like this to unusability from the outside. Although if they embed a cellular modem in it, all bets are off.
I use RBLs here and we block about 35% of incoming as Spam. Given I still get a lot of spam anyway, I'm assuming 40% is probably real accurate.
IIRC they test newborns to see if they are "PK babies" right away, since they'll die before they ever get their first diet coke due to phenylalanine exposure in regular foods.
My friends from the fringe are constantly chiding my use of Nutrasweet, but honestly I like it better than sugar and it doesn't make me feel all weird like lots of sugar does. The only other sweeteners I use besides nutrasweet are honey and molases, but these are flavorizers as much as sweeteners.
I just wish nutrasweet was more shelf-stable and high temperature stable so I could substitute it for sugar in recipes and not have the diet coke in the back of the fridge go south.
Once every few years? I upgrade more than that, at least to keep up with the bind bugs and sendmail bugs. Yes, I could binary update those specific bits with patches, but there's nothing to be lost with a buildworld in my experience (4.0 through 4.8-RC), and it keeps everything coherent. It seems like keeping up is easier if you at least follow the minor point updates to stable as well, as occasionally something interesting gets MFC'd requiring a little more work.
The buildworld process is about as unsupervised as a Win2k service pack. There's a few build commands, a mergemaster run, and then a reboot. It doesn't require total attention, but its not 100% background either. And once in a great while it breaks, requiring some investigating to see what's broken (me or world).
Perhaps braver people than I just do binary updates for bug fixes and only rebuild world every few years, but I like it more frequently than that. We usually do enough hardware updates that there's some motivation at least every 18 months to swap around boxes, which usually involves at least some kind build if not a total reinstall.
So it's pretty much total BS, then?
My experience managing Win2k (12 servers, 3 locations) is that it's really quite simple to manage and even simpler to keep updated. It helps to have high speed (256k, minimum) connectivity to the sites for terminal services (reboots, service packs, etc).
A buildworld on one of my FreeBSD boxes takes much longer than even a Win2K SP with backup, plus there's time spent re-building stuff from ports when that needs updating.
I'd say they're roughly equal in terms of time; FreeBSD needs updating less often, but takes longer and is more involved. Win2k is quicker to update, but you have to do it more often.
If *ALL* I did was manage the boxes (OS updates, hardware maintenance, upgrades), I could do an equal number of Win2k and FreeBSD machines, and probably 2-3 times what I do now. But I also have to do all network infrastructure (routers, switches, dialup, cabling) as well as deal with the politics, PHBs and end users, all of which sucks valuable time.
I'm not a huge fan of traditional multidesktop environments because they're too geographically distinct -- I have to "switch" between them, and that involves an annoying redraw penalty as well as a mental "reset" to refamiliarize myself with that environment.
I'm not sure get the 3D UI completely, but what would make sense to me would be putting the user inside of a many-sided polygon container. Each interior face of the sphere would be a typical desktop. The controls for the GUI would enable zooming out to see multiple faces/desktops simultaneously, as well as the ability to pan over desktops, in addition to being able to "lock" onto a desktop. Windows could be moved between desktops all-at-once, or dragged if you were panned over them.
It would have interesting multimonitor potential, as well -- the monitors could be assigned as geographically adjacent so that they showed adjacent polygonal surfaces/desktops, the polygonal surface could be sized to the combined monitor resolution or the monitors could be 'detatched' so that they showed different views (perhaps one zoomed out, one locked).
I don't think this would be that hard to implement, either (disclaimer: not a developer), since it wouldn't involve changes in the base GUI, just in the way that desktops as a whole are presented and navigated.
Wow, that's a huge freaking rocket. It looks about the same size as the ones you'd see on mobile launchers in the old red square parades.