Actually *I* am, having an FCC license that allows me to modify and experiment with transmitters -- and holding ME responsible that they meet FCC regs.
Maybe Buffalo submitted their gear and got it approved.
The FCC does care what modifications are performed by end users on licensed devices. these devices are licensed, by the FCC as 'part 15' devices which allow NO modifications of the RADIO by end users (not even extended antennas). they must be used exactly as licensed. modifications of the non-radio portions are not a problem. broadcom and other provide 'blobs', in part, to comply with this. If you modify the radio, by putting on a big transmitter say, if the FCC catches you(big if) you will be subject to civil and/or criminal penalties. Based on research i did to repurpose one of these i suspect the following:
broadcom radios and ethernet are weird, and so the drive has no access to the hardware per se, or even to a hardware interface as we expect it. rather the 'blob' represents the executable code for several processors (perhaps 4 or more) on the interface card. it is downloaded to the interface at initialization time. the interface cpus talk to hardware at a very low level/hardware specific manner and timing is important, foul up the timing and you can be transmitting on a different band or with unacceptable distortion. I believe there are several older mips cpus controlling various bits of the radio and directly generating various waveforms needed in the radio. the kernel drivers essentially IPC to one of the cpus on the interface, and it talks to the others, or modifies the control structures they use directly.
even users with special FCC licenses that allow development of devices for other uses under other parts of the
FCC regulations (re-purposing 'commercial off the shelf' equipment) can not get the code, i have been told its really a mess and not easily modified, small local changes can changed timings and effect other things. Ubiquity provides better support. Atheros provides some support, and a more reasonable hardware interface.
how 'bout this though..
hot-swap mirrored drives. pull 1/2 of the mirror at any time to make a backup. replace the pulled drives with blanks.
keep a short stream of backup drives, say 8 or 9. drives are cheap. store in well padded metal boxes, offsite.
Let me see now, in order for this to work each user will need the tools and data that they neeed/use any time at all... all provided by the ISP. Web apps, well..no.. imagine autocad, or frame, or my custom cost accounting application, available from the ISPs. Heck I need three version of Java to run applications i currently need.
How will the software vendors make any money? How will the ISPs charge for these things? For this to really work
o The ISPs will provide all of the apps, in all
the correct versions for all of their
customers.
o They will provvide tools to import all of the
reams of data (in a variaty of formats) to the
apps they support
o They will provide the support they I require
in a timely fashion (right from cable
companies and telcoms...)
No, most of the problems being discussed are startup problems. Systems are still immature and difficult to take care of and the population in general has a high level of miss understanding of whats happening. The young people just starting high school have a different perspecive from my mother and sister (who's systems I administer).
This remindes me of a comment i read in a 1919 magzine, something about cars not being useful and all the people would be moving onto buses and using them like trains, because most people do not want to learn to be a mechanic, and horses are more frendly and less trouble and...
And like my mom does with her car, professional support will solve the problem for many if not most people.
I am a private pilot and HAVE had a tape recorder/player in the cockpit cause significant interference on a navigation insturment in the cockpit. Durring a flight from San Luis Obispo to Santa Barbara the navi reciever indicated an errror when homing to the Gaviota VOR. The error was about 15 deg. The conditions were vfr and there was no safty-of-flight problem.
the real issue, i suspect, is RFI and the amount of intercell interference.
There are only 329 frequencies in the low (VHF) cell band and some 500 or so in the upper (UHF) band. No cell can use the same freq as another cell within some specified distance. Because of this the avarage cell can only use a small number of them (I remember somehing like less than 40 for the vhf band) to prevent RFI.
the cell companies hire engineering firms to calculate the interference and calculate cell sizes, antena siting, and transmitter power (at the base station). Many areas are saturated, the maximum number of cells are installed with the lowest power transmiters and directiona antanas. the cells in a region (for a carrier, which is allocated a set of frequencies) communicate and decide what cell gets what frequencies now.
A cell phone in an airplane will blank that frequency from all of the cells that it is detected in. So some 330 calls from aircraft over san franciusco, for instance, could block all the vhf (old style analog phone) in the whole bay area!
the same holdes for the newer phones, but with the fancier multiplexing schemes used the calculations are more difficult and probablistic (ie: 650 calls at once have a 50% chance of blocking all the UHF/digital cells in the above case).
Currently there are export restrictions on this kind of thing, if done in USA or European country. I suggest alternative.
We have small floating nation/ship with no such restrictions. Only 3/4 done it is under construction in S. Korean ship yard, for launch in 2006 or 2007. We whould have no sush restrictions.
In fact, we may be willing to be a beta customer, and even help with the initial product launch.
The shuttles are fine, but still have 150 to 200 watt pwr supplies. there are other options, check out:
http://www.logisysus.com/cbox.htm
for mobile systems. some of these are designed to run on batteries (12v pwr supply) and have 90 or 120 watt pwr supplies. I hjave no connection to these people and have not done bussiness with them. just discovered them looking for computers for a sailboat.
multi master data-bases do updates on all the masters at the same time using a two-phase (or other exotic) commit mechinism. All of the masters are updates in the same transaction. Kinda ups the bandwidth requirements on the backbone between the masters. To avoid the reliablity problem, down servers are noted and updated later (before they come back online) by log sniffing or something.
Well clusters of type 1 don't really implement any redundency at all. A failure of any node in the system, makes some of the data unavailable. So this might not be the best for the original posters problem. It does allow a high degree of scaleablity in terms of database size and performance (assuming you solve the moves and changes problem). In fact if access to all of the data is a requirement, the more nodes in a type 1 (shared nothing) cluster the lower the reliablity.
Well not only is tranactionality a problem at the time of a failure, it impacts recovery too.
When a failed node is restarted, it must be brought up to date and the initial replication structure restartd, without loosing anyting. Additionaly a falure must be allowed durring this process.
It gets complicated, and any solution based on rsync is not a solution.
I the problem i have with this is step 7.
7. Criminal punishment should apply where the wrongdoing adversely affects the general public's interests. Criminal authorities represent the public's interests.
The problem with this is the merkeyness of the definition of what adversly affects the general public's intrests. At some level someone can argue that ANYTHING adversely effects the public intrests. There is no clean test of this so anything or nearly anything can be construed this way. These merky 'lines' lead to ever expanding definition of public intrests. This IS happening with copygight law.
Didn't any one read the article on pricing about two months ago? No one but us suckers (customers) think prices should be related to costs. Charge what the market will bare.
There are many reasons for this, from a lack of study to inattention on the part of programers. Another reason is change. Code changes unexpected ways and breaks other things. Then there is lack of testing and inadaquate unit test facilites. Or the failure to use these often enough when things change.
One problem is that IDEs and debuggers are good tools, and often the only one provided with compilers, so we use them. They are not the only kind of tool available. They are not the best tool for all bugs either.
Other kinds of tools are available, usefull and ignored. Code coverage tools, built in trace facilities and other tools are useful in creating programs with few errors.
Another issues is that we (as programmers) design fragile facilities. When other things change, these produce wrong results, that show as bugs much latter. Where I work we use raw use counts in several places. Bugs in the code that uses this are very hard to find. With a little thinking this can be made much more robust, and easier to debug.
This is just an example and there are many other similar opertunities. Opertunities tomake the code more robust and to catch its own errors. Many conventional C/C++ facilities suffer from this. Malloc is a common example, are are points, and other things.
here are also systemantic issues with the code/ runtime enviornment in various languages. These are large for c/c++ programmers, were these languages make it easy to damage the runtime enviornment. Other languages like Java and basic make it harder.
In any case, developeres can put defensive features arround the fragil facilities to improve debugging and reliability. These all cost execution time and memory, but as the original poster mentioned, these are more available. IN any case properly designed facilites and dissapear or reduce in production versions.
Another problem is knowing when and what to test. You can test forever and still have buggy code, if you dont do the right tests. Testing the right code and the right cases is a whole problem in its self. How do you know that you test plan is adaquate? Coverage analysis is a good start, but few developers use it.
Well my experience in high tech is there are't enough american born and trained engineers. Period. To deny foreign engineers a job in the US it to ensure that the work that the do will be done elsewhere. I have held jobs open for almost a year and not gotten a single qualified US born engineer to apply. And I cant spend 5 or 6 years to train a beginner. Thats what college is for.
Big multinational companies can arange to do the work where ever is handy. Outsourcing is simple and a fact of life. H1B visas are just a way to allow the high tech industry a way to grow faster than the educational system can provide grads. Companies can move the jobs overseas, and will if the alternative is dropping services. The state help desk in India is a good example, you choose which you would prefer:
1) move the help desk back to the US
2) cut medical benefits in the state or
something or the cost of education
3) raise taxes?
In another sense H1B visas are the only way that we remain shielded from the poor quality of many our schools and the limited output of the better ones. At one point durring the high tech boom, in the bay area there were more unfilled high tech jobs than the entire number of people that would graduate, the next year, with approprate degrees (or so it was reported in the local news rags). I beleve it, i could not heir qualified people. It allows us to import the best and brightest from overseas.
there are surely abuses and problems with the H1 program and they should be fixed, but theis seems like a way to guarentee that more and more high tech work moves overseas.
One of the problems that i find with GUIs is that they make an assumption about 'the user'. There is no 'the user', there is a community of users. All have differing desires, requirements, style etc. and the GUI designers pick one, or a small set of them and make a GUI.
This is sure to be wrong for the majority of the community of user. And particularly for anyone that resembles a power user, as, as another poster mentioned, they alway include the clueless in 'the user' they design for.
The GUI give pleasure in different ways for different members of the user community. As such some people fit the available GUIs better than others. My sister (non-tech artist) finds Win and Mac pleaurable as both empower her, and allow here to get on with the things that she does.
For me a different approach works better.
The problem is one of makeing one-size-fit-all or
taking customizablity to the next level. GUIs should stop making users adapt to them, but rather adapt to the users. Not just in look and feel, but in behavor and action too.
There are many apps that are still memory and compute bound. Voice recognition could easly use 4x more memory and more compute power, other apps (decode hd tv, monitoring video, home security) can all take big bits. Heck even the fridg that reads bar codes needs a place to store the UPC and usage data, and you probably want a file server, rather than the hard disk in the fridg.
Well there is at least one more, common in corprate development enviornments:
A large team (of hundreds or thousands of individuals) with differing backgrounds (not all programers) working on a complex project at once. While each may have seperate 'project' they all impact the same code base. This is really the problem the Clearcase, CCC and some of the other larger CM systems are solving; keeping 1000 people out of each others hair.
I have worked on several projects with 400 developers, graphic artists (defineing graphics), online documentation (integrated with the GUIs), etc. working on the same code base.
To do this well you need atomic operations (most CM fails here), branching (can ever break the nightly build or half the people stop working), good merging (will often have merge conflicts), profile/view management (allows you to fix bugs in the release and work on the current stuff), various kinds of rules, and distributed operations at several levels.
as a long time Clearcase user i agree. it was very powerfull, and dificult to use. 10 or 12 people supported our development community.
They managed branches, cross site syncs, and made scripts to do almost everything. The GUIs don't allow user scripting, and its really needed here.
Without this support users made MANY mistakes when using Clearcase for the first three months.
Had a GOOD merge utility though. We had hundreds of branches.
Slashdot Mirror
Actually *I* am, having an FCC license that allows me to modify and experiment with transmitters -- and holding ME responsible that they meet FCC regs.
Maybe Buffalo submitted their gear and got it approved.
The FCC does care what modifications are performed by end users on licensed devices. these devices are licensed, by the FCC as 'part 15' devices which allow NO modifications of the RADIO by end users (not even extended antennas). they must be used exactly as licensed. modifications of the non-radio portions are not a problem. broadcom and other provide 'blobs', in part, to comply with this. If you modify the radio, by putting on a big transmitter say, if the FCC catches you(big if) you will be subject to civil and/or criminal penalties. Based on research i did to repurpose one of these i suspect the following:
broadcom radios and ethernet are weird, and so the drive has no access to the hardware per se, or even to a hardware interface as we expect it. rather the 'blob' represents the executable code for several processors (perhaps 4 or more) on the interface card. it is downloaded to the interface at initialization time. the interface cpus talk to hardware at a very low level/hardware specific manner and timing is important, foul up the timing and you can be transmitting on a different band or with unacceptable distortion. I believe there are several older mips cpus controlling various bits of the radio and directly generating various waveforms needed in the radio. the kernel drivers essentially IPC to one of the cpus on the interface, and it talks to the others, or modifies the control structures they use directly.
even users with special FCC licenses that allow development of devices for other uses under other parts of the FCC regulations (re-purposing 'commercial off the shelf' equipment) can not get the code, i have been told its really a mess and not easily modified, small local changes can changed timings and effect other things. Ubiquity provides better support. Atheros provides some support, and a more reasonable hardware interface.
redundant power supplies?? these are server class machines, and i would expect high reliability.
how 'bout this though.. hot-swap mirrored drives. pull 1/2 of the mirror at any time to make a backup. replace the pulled drives with blanks. keep a short stream of backup drives, say 8 or 9. drives are cheap. store in well padded metal boxes, offsite.
I wonder how bright a 1.4 watt laser will be when spread over a wall from 30 ft. Or does this only project a monitor sized immage at 2 ft??
This makes a lot of sense..
Let me see now, in order for this to work each user will need the tools and data that they neeed/use any time at all... all provided by the ISP. Web apps, well..no.. imagine autocad, or frame, or my custom cost accounting application, available from the ISPs. Heck I need three version of Java to run applications i currently need.
How will the software vendors make any money? How will the ISPs charge for these things? For this to really work
o The ISPs will provide all of the apps, in all
the correct versions for all of their
customers.
o They will provvide tools to import all of the
reams of data (in a variaty of formats) to the
apps they support
o They will provide the support they I require
in a timely fashion (right from cable
companies and telcoms...)
No, most of the problems being discussed are startup problems. Systems are still immature and difficult to take care of and the population in general has a high level of miss understanding of whats happening. The young people just starting high school have a different perspecive from my mother and sister (who's systems I administer).
This remindes me of a comment i read in a 1919 magzine, something about cars not being useful and all the people would be moving onto buses and using them like trains, because most people do not want to learn to be a mechanic, and horses are more frendly and less trouble and...
And like my mom does with her car, professional support will solve the problem for many if not most people.
I am a private pilot and HAVE had a tape recorder/player in the cockpit cause significant interference on a navigation insturment in the cockpit. Durring a flight from San Luis Obispo to Santa Barbara the navi reciever indicated an errror when homing to the Gaviota VOR. The error was about 15 deg. The conditions were vfr and there was no safty-of-flight problem.
the real issue, i suspect, is RFI and the amount of intercell interference.
There are only 329 frequencies in the low (VHF) cell band and some 500 or so in the upper (UHF) band. No cell can use the same freq as another cell within some specified distance. Because of this the avarage cell can only use a small number of them (I remember somehing like less than 40 for the vhf band) to prevent RFI.
the cell companies hire engineering firms to calculate the interference and calculate cell sizes, antena siting, and transmitter power (at the base station). Many areas are saturated, the maximum number of cells are installed with the lowest power transmiters and directiona antanas. the cells in a region (for a carrier, which is allocated a set of frequencies) communicate and decide what cell gets what frequencies now.
A cell phone in an airplane will blank that frequency from all of the cells that it is detected in. So some 330 calls from aircraft over san franciusco, for instance, could block all the vhf (old style analog phone) in the whole bay area!
the same holdes for the newer phones, but with the fancier multiplexing schemes used the calculations are more difficult and probablistic (ie: 650 calls at once have a 50% chance of blocking all the UHF/digital cells in the above case).
Currently there are export restrictions on this kind of thing, if done in USA or European country. I suggest alternative. We have small floating nation/ship with no such restrictions. Only 3/4 done it is under construction in S. Korean ship yard, for launch in 2006 or 2007. We whould have no sush restrictions. In fact, we may be willing to be a beta customer, and even help with the initial product launch.
The shuttles are fine, but still have 150 to 200 watt pwr supplies. there are other options, check out: http://www.logisysus.com/cbox.htm for mobile systems. some of these are designed to run on batteries (12v pwr supply) and have 90 or 120 watt pwr supplies. I hjave no connection to these people and have not done bussiness with them. just discovered them looking for computers for a sailboat.
multi master data-bases do updates on all the masters at the same time using a two-phase (or other exotic) commit mechinism. All of the masters are updates in the same transaction. Kinda ups the bandwidth requirements on the backbone between the masters. To avoid the reliablity problem, down servers are noted and updated later (before they come back online) by log sniffing or something.
Well clusters of type 1 don't really implement any redundency at all. A failure of any node in the system, makes some of the data unavailable. So this might not be the best for the original posters problem. It does allow a high degree of scaleablity in terms of database size and performance (assuming you solve the moves and changes problem). In fact if access to all of the data is a requirement, the more nodes in a type 1 (shared nothing) cluster the lower the reliablity.
This works fine until a disk system fails or a drive crashes. RAID is somewhat of an aid here, but all RAID systems have some single point of failure.
Well not only is tranactionality a problem at the time of a failure, it impacts recovery too. When a failed node is restarted, it must be brought up to date and the initial replication structure restartd, without loosing anyting. Additionaly a falure must be allowed durring this process. It gets complicated, and any solution based on rsync is not a solution.
Do we really want our voting machine to record some sort of voter id (to ensure uniqueness) and our vote?
Its not a description, its a name. Just like i am ba_hiker. It dosen't tell anything about me, IP dosen't tell you what the concept it names is about
I the problem i have with this is step 7. 7. Criminal punishment should apply where the wrongdoing adversely affects the general public's interests. Criminal authorities represent the public's interests. The problem with this is the merkeyness of the definition of what adversly affects the general public's intrests. At some level someone can argue that ANYTHING adversely effects the public intrests. There is no clean test of this so anything or nearly anything can be construed this way. These merky 'lines' lead to ever expanding definition of public intrests. This IS happening with copygight law.
Didn't any one read the article on pricing about two months ago? No one but us suckers (customers) think prices should be related to costs. Charge what the market will bare.
There are many reasons for this, from a lack of study to inattention on the part of programers. Another reason is change. Code changes unexpected ways and breaks other things. Then there is lack of testing and inadaquate unit test facilites. Or the failure to use these often enough when things change.
One problem is that IDEs and debuggers are good tools, and often the only one provided with compilers, so we use them. They are not the only kind of tool available. They are not the best tool for all bugs either.
Other kinds of tools are available, usefull and ignored. Code coverage tools, built in trace facilities and other tools are useful in creating programs with few errors.
Another issues is that we (as programmers) design fragile facilities. When other things change, these produce wrong results, that show as bugs much latter. Where I work we use raw use counts in several places. Bugs in the code that uses this are very hard to find. With a little thinking this can be made much more robust, and easier to debug.
This is just an example and there are many other similar opertunities. Opertunities tomake the code more robust and to catch its own errors. Many conventional C/C++ facilities suffer from this. Malloc is a common example, are are points, and other things.
here are also systemantic issues with the code/ runtime enviornment in various languages. These are large for c/c++ programmers, were these languages make it easy to damage the runtime enviornment. Other languages like Java and basic make it harder.
In any case, developeres can put defensive features arround the fragil facilities to improve debugging and reliability. These all cost execution time and memory, but as the original poster mentioned, these are more available. IN any case properly designed facilites and dissapear or reduce in production versions.
Another problem is knowing when and what to test. You can test forever and still have buggy code, if you dont do the right tests. Testing the right code and the right cases is a whole problem in its self. How do you know that you test plan is adaquate? Coverage analysis is a good start, but few developers use it.
Well my experience in high tech is there are't enough american born and trained engineers. Period. To deny foreign engineers a job in the US it to ensure that the work that the do will be done elsewhere. I have held jobs open for almost a year and not gotten a single qualified US born engineer to apply. And I cant spend 5 or 6 years to train a beginner. Thats what college is for. Big multinational companies can arange to do the work where ever is handy. Outsourcing is simple and a fact of life. H1B visas are just a way to allow the high tech industry a way to grow faster than the educational system can provide grads. Companies can move the jobs overseas, and will if the alternative is dropping services. The state help desk in India is a good example, you choose which you would prefer: 1) move the help desk back to the US 2) cut medical benefits in the state or something or the cost of education 3) raise taxes? In another sense H1B visas are the only way that we remain shielded from the poor quality of many our schools and the limited output of the better ones. At one point durring the high tech boom, in the bay area there were more unfilled high tech jobs than the entire number of people that would graduate, the next year, with approprate degrees (or so it was reported in the local news rags). I beleve it, i could not heir qualified people. It allows us to import the best and brightest from overseas. there are surely abuses and problems with the H1 program and they should be fixed, but theis seems like a way to guarentee that more and more high tech work moves overseas.
Well this is fair, and so is us ISPs charging indian ISPs to access US content.
One of the problems that i find with GUIs is that they make an assumption about 'the user'. There is no 'the user', there is a community of users. All have differing desires, requirements, style etc. and the GUI designers pick one, or a small set of them and make a GUI. This is sure to be wrong for the majority of the community of user. And particularly for anyone that resembles a power user, as, as another poster mentioned, they alway include the clueless in 'the user' they design for. The GUI give pleasure in different ways for different members of the user community. As such some people fit the available GUIs better than others. My sister (non-tech artist) finds Win and Mac pleaurable as both empower her, and allow here to get on with the things that she does. For me a different approach works better. The problem is one of makeing one-size-fit-all or taking customizablity to the next level. GUIs should stop making users adapt to them, but rather adapt to the users. Not just in look and feel, but in behavor and action too.
There are many apps that are still memory and compute bound. Voice recognition could easly use 4x more memory and more compute power, other apps (decode hd tv, monitoring video, home security) can all take big bits. Heck even the fridg that reads bar codes needs a place to store the UPC and usage data, and you probably want a file server, rather than the hard disk in the fridg.
Well there is at least one more, common in corprate development enviornments: A large team (of hundreds or thousands of individuals) with differing backgrounds (not all programers) working on a complex project at once. While each may have seperate 'project' they all impact the same code base. This is really the problem the Clearcase, CCC and some of the other larger CM systems are solving; keeping 1000 people out of each others hair. I have worked on several projects with 400 developers, graphic artists (defineing graphics), online documentation (integrated with the GUIs), etc. working on the same code base. To do this well you need atomic operations (most CM fails here), branching (can ever break the nightly build or half the people stop working), good merging (will often have merge conflicts), profile/view management (allows you to fix bugs in the release and work on the current stuff), various kinds of rules, and distributed operations at several levels.
as a long time Clearcase user i agree. it was very powerfull, and dificult to use. 10 or 12 people supported our development community. They managed branches, cross site syncs, and made scripts to do almost everything. The GUIs don't allow user scripting, and its really needed here. Without this support users made MANY mistakes when using Clearcase for the first three months. Had a GOOD merge utility though. We had hundreds of branches.