They'd have to save the state of the whole machine on every watchdog tick. Hashing the whole memory is out of question, you'd need to hash just pages written to since the last tick, and that requires setting up page faults on the rest (or some custom architecture with generational memory, also costly).
Actually, the authors extend LLVM and statically instrument code so they only incur overhead when actually checking for infinite loops, and it is fairly lightweight since it can take advantage of this static knowledge to only record the state that is relevant to the loop continuing or exiting..
Second, a microkernel architecture, while more robust than a monokernel, cannot be as fast as the monokernel. If one subsystem wants another subsystem to do something, it must format and send a message; the other subsystem then receives the message, unpacks it, validates it, and then does the action. This is more secure and more stable than the monokernel, where the one subsystem will just make a function call in the other subsystem's code; but it is inherently slower.
Actually, this isn't really true. If the TLB were tagged with a process ID it would not need flushing on every context switch (MIPS used to be like this). On a modern machine that is going to be the biggest hit you take. There is also message copying--an async IPC kernel like Mach can never win here, but L4 generally copies no data and can perform an IPC with only one or two context switches. The small size of a modern microkernel means you get much better cache usage and so in theory can be faster.
Of course, GNUMach based Hurd is doomed. But it's not because of fundamental problems with microkernels--just Mach (which is understandable since it the first generation... you have to learn from your mistakes somehow).
I leave my mail on the IMAP server of the hosting cooperative I'm in (for easy access anywhere, they keep backups so I don't have to, etc), and it looks like I unknowingly abandoned the data! I trust the cooperative infinitely more than I would trust e.g. gmail, but it looks like law enforcement can just as easily request access to my old mail there.
I'd argue instead that the machine being imperative is the problem. It's 2011 and we still just have a gigantic array of words that we can move an instruction pointer through... See e.g. Lambda: The Ultimate Opcode for an alternative approach to system design.
Nothing really prevents us from building an expression reduction machine (with an imperative IO processor, naturally... it is nice to interact with the external environment) nowadays. I'm not sure where the rest of the project went, but there was a project involving Ivan Sutherland at Berkeley that built a purely parallel processor. In FLEET there is no notion of a linear program nor of linearly addressable memory--programs become dependency graphs that are reduced.
Basically, the machine being a really fancy Turing machine is an accident (in the formal sense), and not at all essential to computation. It happens to work pretty well and certainly will be dominant for a long time, but forever? Probably not, expression reduction machines can be clockless, massively parallel (as in executing thousands of operations simultaneously), &c -- which means smaller dies, lower power usage,....
Another thing to think about is that this is for freshman. I'm suspecting that anyone who can get into CMU has a pretty good grasp on Algebra... and probably the differential/integral calculus. Functional programming is a natural extension of the mathematics the students have already learned whereas OO is an entirely different paradigm which presents a barrier to actually learning the essence of computing science.
That, and CMU is the Land of SML. The mantra there is that OO is a hack--an artificial conglomeration of several orthogonal concepts: modules, protocols, records, polymorphic functions, &c. Under that philosophy OO is just another obsolete intermediate step in the evolution of computer language.
Part of the issue here is terminology--throwing around CDMA vs GSM instead of CDMA vs TDMA.
Old style GSM used TDMA, the GSM 3G specifications all use CDMA. The protocols running on the layer above how signals are multiplexed are still different. GSM TDMA networks may still be running, but really the faster they die the better, GSM TDMA is a horrible mess (there's a lot of good info over at Harald Welte's Web Log) that is incredibly difficult to scale.
Interestingly enough, EV-DO is now a 3GPP standard and so... technically all of the US carriers are GSM providers now. Nothing except the will to remain the sole provider of devices prevents Verizon/Sprint from giving users USIMs at this point...
Since Verizon et al finally joined the 3GPP, there was actual work toward unifying the myriad network and radio protocols that have accreted over the last 25 years of digital wireless, and this is LTE. It looks like all Northern American providers at least want to ditch their old tower equipment and move as quickly as possible toward this.
There is some hope that Verizon and friends will finally start offering phones with USIMs and play nicely with the rest of the world, but I somehow doubt that (I am an eternal pessimist). Especially in light of yet another major carrier being swallowed by AT&T II (Cingular/ATTWS)... how long until AT&T I (Verizon, formerly Bell Atlantic...) reswallows AT&T II) finishes the process and that's all we really have left?
Still, from a technical perspective what are traditionally known as CDMA and GSM are indeed merging into one standard, and this is probably a good thing. Hopefully the FCC and FTC will step up and ensure competition (heh).
There is no excuse for a driver to not notice a bicycle hitting him. If a driver were that inattentive then he would be a hazard to everyone on the road and would not deserve a license.
Oh wait, reality calls where almost everyone is like that. This is why being a cyclist is so damn dangerous (hello car making a left turn into me; I am in the intersection and perhaps you would see me if you got off of your phone and looked where you were turning).
When you do not have money it is damn near impossible to afford a car that does not guzzle gas due to having to buy a used vehicle, and then one that you can afford payments on (or to buy in cash).
Sure, it would make more sense to spend $30k on a hybrid or turbodiesel, but when you make $22k a year... I doubt that her SUV is one she got new.
You cannot immerse someone in a craft for a few days and expect him to get a positive experience. Crafts take a long time to master, and working at the master level can be confusing and difficult.
This, however, is not a problem. It is infeasible to simply drop out of your stable job and jump to a completely unrelated field anyway. This, however, works well with learning a craft. It is easy to devote a few hours every weekend to developing a craft skill (e.g. home brewing beer starting with extract, moving to steeping, partial mash, full mash,... while experimenting with recipes). It is easy to start most crafts on a small scale without a huge initial investment, and that lets a person try something out and gradually gain the skill without requiring him to give up his primary income source.
Some people will go on to the full time level, but many will not. Some maintain a certain level of skill in an area (e.g. many home brewers will never go past doing minimashes because they are happy with their skill level as a brewer), and thus gain an enjoyable hobby. At worst it some will find that what they thought would be fun is actually unenjoyable, and a wasted afternoon doing something is better than a wasted afternoon doing nothing.
Now, there are some crafts that are difficult to learn without a lot of equipment and time. Money can't help you there. If you have strange enough friends you can find someone to teach you [and what is life without strange friends?].
You don't need millions of dollars to make a fairly low volume graphics card ASIC. You can have your screens fabricated by contract. OGP also has the Open Graphics Development (OGD) board. It seems to be a nice and powerful FPGA board (with plenty of GPIO for other projects) for a very low price (compared to every other FPGA board on the market). I think that alone may be enough to keep them afloat.
And of course, even if it does fail, they tried. It is a noble thing to attempt this, and not a waste of time. At the very least they will all walk away with a few interesting years of life.
I speak of Freedom and not price. You are free to modify the system however you like, but the sad reality is that sometimes people must be forced to share or they will be leaches and steal. Thus the GPL forces you to share your work. It is sad that people must be forced to do so, but history has shown that there are a large number of people who will exploit the kindness of others for their own gain. This is not about profit, but about sharing knowledge to further the human race as a whole.
The only freedom the GPL doesn't give you is the freedom to steal. I think that is quite reasonable.
I do not represent Open Source. I hate Open Source.
Short sightedness is the cause of most humanity's problems. It is a shame that most people are like you, and fail to care about the damage you are doing. Buying nVidia is rewarding anti-social behaviour ("we can't share they are OUR ideas not yours").
They said the same things about the first versions of GCC (so far behind the proprietary compilers), Linux (so far behind the proprietary kernels), and pretty much all Free Software.
I will be buying one. I think that there are just enough people to keep the project alive so they can develop a second faster card.
Please, I hope that is does scare people like you away. I run GNU/Linux because I am free to modify every part of my system. A 'Free' system that relies on proprietary drivers is no longer Free. Users like you help to destroy my freedom.
You can just go back to using Windows or OS X. You are not welcome in our community (what a crazy idea! A community where everyone contributes!).
Those systems have PCI, and could use the cards. But they can't anymore because of the x86-only drivers.
What if the driver is unstable on SMP and you are not in the target market for the developer so you can't even make it work with SMP? Hey, you just got fucked out of $NNN for a new card that you can't use unless you want to disable your second CPU.
What if you are doing kernel dev, or just want to test some of the newish stuff like 4kB stacks or regparm argument passing? Out of luck yet again.
What if you are working on a new graphics system, and don't want to use X? Well, with nVidia you are out of luck, and with ATi you are only able to do it if you base your system or DRI (and even then, a lot of times it won't work).
No, prison management companies receive a lot of money and then pocket it while treating the prisoners like shit (doing things like keeping medication from sick prisoners).
Remember that punishment has been shown to not rehabilitate people anyway. It merely makes them try harder to not get caught (in most cases).
How the Hell can you suggest that we cut back on food and heating for prisoners? They are still people, and are already being treated as animals (Sodexho...)
If I need to stay up a few extra hours I have found that a pot or two of tea helps greatly. The energy increase is slight and prolonged instead of intense and quickly fading. All of the sugar in the energy drinks makes me sleepy (something with insulin and whatever that makes you tired after a big meal). A nice Yuannin/Darjeeling blend has become my standard for all nighters. As soon as the first woody sip I am reminded of nights past...
The real way to get more energy is to make sure you are taking vitamins (it is difficult to eat a really balanced diet), and do a small amount of exercise. A few pushups and situps in the morning, and a bit of walking during the day do wonders (I have the advantage of going to college, so I can wander around campus between classes). I have found that waking up an hour early, and then spending the hour drinking a pot of tea and breakfast also makes the day better. As tempting as hitting snooze is, putting my shower off and doing things that require no mental or physical effort (put water in the kettle, weigh the leaves, drink a glass of water, toss a bagel in the toaster and preheat the pot right before the kettle boils, set the timer for four minutes and let the tea steep, add cream cheese to the bagel, eat and drink). It is a good way to gradually wake up (the caffeine helps too).
As for why I (*shudder*) at the thought that students are responsible for some of the code in BSD (and also Linux, I shudder at that too), I have seen the code that students write. I've seen this come out of a student's garbled mind (written in C)...
Which students? The stupid ones who shouldn't be in comp sci? Those kind don't actually enjoy coding, and are not the ones who work on GNU/Linux/BSD. The good comp sci students, however, produce good code because they enjoy coding. Remember who was a college student with Linux was first written?
And there are other parts that make it sound like LiveJournal would never be in trouble for this unauthorized access parts. But really, who would bother to post their thoughts and words on a site that has no garauntee of saving them? At any minute, LiveJournal could format its servers and databases and start over with no one able to say anything.
Using a client like Logjam (or emacs lj-update) you can synchronize your journal with an offline copy. The LJ code base is Free Software (well, most of it anyway) so you could always set up your own LJ server. The local entries in logjam are just a directory tree of xml files (with each xml file being a month of entries in a very simple xml format), and it wouldn't be difficult to write a program to translate them into any format you wanted (I could do this fairly quickly using Scheme + SSAX). The only issue left would be external images. I store all of my lj related images on my site in a special dir so I would still have those, and you can now insert images from Scrapbook into your journal so someone could write a ScrapBook backup script, and then you could write a quick perl script to find/replace all of the scrapbook references with whatever host you wanted to copy the images to.
As long as you keep a local backup of your journal (which is easy to do with tools easily available for any system) it would be fairly painless to move them anywhere else. The tasks are simple enough for even a not-so-good programmer to code up into a simple application for others to use.
The amount of pressure needed to floor the gas is only slightly higher than that need to move forward a 1 MPH. Likewise, unless you're driving a huge 40 ton earthmover, braking force to lock the wheels is only slightly greater than the force necessary to gently stop the car.
My 91 Camaro disagrees with you. The pedal has low resistance until around halfway down when the resistance gets progressively larger until it hits the floor. It isn't much, but it is far stiffer than most cars I have driven (it is really strange to drive my car after driving another one for more than a day or two). It isn't so much that you have to stomp, but the force needed to push the pedal past halfway it not insignificant (and yes, I do lubricate the pedal on the schedule my service manual says to).
I use ffmpeg to transcode episodes of TV shows for my phone (Nokia 7610). They run about 30M per episode, and are surprisingly watchable. The only thing I wished is that I had the Nokia stereo headphones (even though the 7610 only does mono it still outputs to both ears) if only so that I wouldn't have to have one ear open. It helps a lot to kill the four hours between classes I have on Tuesday and Thursday. Having a 512M RS-MMC card comes in handy.
Even if I transcode them at 35kbps (video + audio) they are OK to watch. You can stream that over standard GPRS. At 96kbps they get pretty high quality (keep in mind this is QCIF so the video is really low res, but the screen is too so you don't notice). That can be streamed over EDGE fine.
Your mum's recipies would be a great application of LaTeX. Just hack together a recipe style file, and some macros to make entering recipies easier. Then you could print a nicely formatted book of her recipies without much trouble.
Slashdot Mirror
They'd have to save the state of the whole machine on every watchdog tick. Hashing the whole memory is out of question, you'd need to hash just pages written to since the last tick, and that requires setting up page faults on the rest (or some custom architecture with generational memory, also costly).
Actually, the authors extend LLVM and statically instrument code so they only incur overhead when actually checking for infinite loops, and it is fairly lightweight since it can take advantage of this static knowledge to only record the state that is relevant to the loop continuing or exiting..
Second, a microkernel architecture, while more robust than a monokernel, cannot be as fast as the monokernel. If one subsystem wants another subsystem to do something, it must format and send a message; the other subsystem then receives the message, unpacks it, validates it, and then does the action. This is more secure and more stable than the monokernel, where the one subsystem will just make a function call in the other subsystem's code; but it is inherently slower.
Actually, this isn't really true. If the TLB were tagged with a process ID it would not need flushing on every context switch (MIPS used to be like this). On a modern machine that is going to be the biggest hit you take. There is also message copying--an async IPC kernel like Mach can never win here, but L4 generally copies no data and can perform an IPC with only one or two context switches. The small size of a modern microkernel means you get much better cache usage and so in theory can be faster.
Of course, GNUMach based Hurd is doomed. But it's not because of fundamental problems with microkernels--just Mach (which is understandable since it the first generation... you have to learn from your mistakes somehow).
There goes the market for Lisp programmers.
I leave my mail on the IMAP server of the hosting cooperative I'm in (for easy access anywhere, they keep backups so I don't have to, etc), and it looks like I unknowingly abandoned the data! I trust the cooperative infinitely more than I would trust e.g. gmail, but it looks like law enforcement can just as easily request access to my old mail there.
I'd argue instead that the machine being imperative is the problem. It's 2011 and we still just have a gigantic array of words that we can move an instruction pointer through... See e.g. Lambda: The Ultimate Opcode for an alternative approach to system design.
Nothing really prevents us from building an expression reduction machine (with an imperative IO processor, naturally... it is nice to interact with the external environment) nowadays. I'm not sure where the rest of the project went, but there was a project involving Ivan Sutherland at Berkeley that built a purely parallel processor. In FLEET there is no notion of a linear program nor of linearly addressable memory--programs become dependency graphs that are reduced.
Basically, the machine being a really fancy Turing machine is an accident (in the formal sense), and not at all essential to computation. It happens to work pretty well and certainly will be dominant for a long time, but forever? Probably not, expression reduction machines can be clockless, massively parallel (as in executing thousands of operations simultaneously), &c -- which means smaller dies, lower power usage, ....
Another thing to think about is that this is for freshman. I'm suspecting that anyone who can get into CMU has a pretty good grasp on Algebra... and probably the differential/integral calculus. Functional programming is a natural extension of the mathematics the students have already learned whereas OO is an entirely different paradigm which presents a barrier to actually learning the essence of computing science.
That, and CMU is the Land of SML. The mantra there is that OO is a hack--an artificial conglomeration of several orthogonal concepts: modules, protocols, records, polymorphic functions, &c. Under that philosophy OO is just another obsolete intermediate step in the evolution of computer language.
Part of the issue here is terminology--throwing around CDMA vs GSM instead of CDMA vs TDMA.
Old style GSM used TDMA, the GSM 3G specifications all use CDMA. The protocols running on the layer above how signals are multiplexed are still different. GSM TDMA networks may still be running, but really the faster they die the better, GSM TDMA is a horrible mess (there's a lot of good info over at Harald Welte's Web Log) that is incredibly difficult to scale.
Interestingly enough, EV-DO is now a 3GPP standard and so ... technically all of the US carriers are GSM providers now. Nothing except the will to remain the sole provider of devices prevents Verizon/Sprint from giving users USIMs at this point...
Since Verizon et al finally joined the 3GPP, there was actual work toward unifying the myriad network and radio protocols that have accreted over the last 25 years of digital wireless, and this is LTE. It looks like all Northern American providers at least want to ditch their old tower equipment and move as quickly as possible toward this.
There is some hope that Verizon and friends will finally start offering phones with USIMs and play nicely with the rest of the world, but I somehow doubt that (I am an eternal pessimist). Especially in light of yet another major carrier being swallowed by AT&T II (Cingular/ATTWS)... how long until AT&T I (Verizon, formerly Bell Atlantic...) reswallows AT&T II) finishes the process and that's all we really have left?
Still, from a technical perspective what are traditionally known as CDMA and GSM are indeed merging into one standard, and this is probably a good thing. Hopefully the FCC and FTC will step up and ensure competition (heh).
There is no excuse for a driver to not notice a bicycle hitting him. If a driver were that inattentive then he would be a hazard to everyone on the road and would not deserve a license.
Oh wait, reality calls where almost everyone is like that. This is why being a cyclist is so damn dangerous (hello car making a left turn into me; I am in the intersection and perhaps you would see me if you got off of your phone and looked where you were turning).
When you do not have money it is damn near impossible to afford a car that does not guzzle gas due to having to buy a used vehicle, and then one that you can afford payments on (or to buy in cash).
Sure, it would make more sense to spend $30k on a hybrid or turbodiesel, but when you make $22k a year... I doubt that her SUV is one she got new.
You cannot immerse someone in a craft for a few days and expect him to get a positive experience. Crafts take a long time to master, and working at the master level can be confusing and difficult.
This, however, is not a problem. It is infeasible to simply drop out of your stable job and jump to a completely unrelated field anyway. This, however, works well with learning a craft. It is easy to devote a few hours every weekend to developing a craft skill (e.g. home brewing beer starting with extract, moving to steeping, partial mash, full mash, ... while experimenting with recipes). It is easy to start most crafts on a small scale without a huge initial investment, and that lets a person try something out and gradually gain the skill without requiring him to give up his primary income source.
Some people will go on to the full time level, but many will not. Some maintain a certain level of skill in an area (e.g. many home brewers will never go past doing minimashes because they are happy with their skill level as a brewer), and thus gain an enjoyable hobby. At worst it some will find that what they thought would be fun is actually unenjoyable, and a wasted afternoon doing something is better than a wasted afternoon doing nothing.
Now, there are some crafts that are difficult to learn without a lot of equipment and time. Money can't help you there. If you have strange enough friends you can find someone to teach you [and what is life without strange friends?].
You don't need millions of dollars to make a fairly low volume graphics card ASIC. You can have your screens fabricated by contract. OGP also has the Open Graphics Development (OGD) board. It seems to be a nice and powerful FPGA board (with plenty of GPIO for other projects) for a very low price (compared to every other FPGA board on the market). I think that alone may be enough to keep them afloat.
And of course, even if it does fail, they tried. It is a noble thing to attempt this, and not a waste of time. At the very least they will all walk away with a few interesting years of life.
I speak of Freedom and not price. You are free to modify the system however you like, but the sad reality is that sometimes people must be forced to share or they will be leaches and steal. Thus the GPL forces you to share your work. It is sad that people must be forced to do so, but history has shown that there are a large number of people who will exploit the kindness of others for their own gain. This is not about profit, but about sharing knowledge to further the human race as a whole.
The only freedom the GPL doesn't give you is the freedom to steal. I think that is quite reasonable.
I do not represent Open Source. I hate Open Source.
Short sightedness is the cause of most humanity's problems. It is a shame that most people are like you, and fail to care about the damage you are doing. Buying nVidia is rewarding anti-social behaviour ("we can't share they are OUR ideas not yours").
They said the same things about the first versions of GCC (so far behind the proprietary compilers), Linux (so far behind the proprietary kernels), and pretty much all Free Software.
I will be buying one. I think that there are just enough people to keep the project alive so they can develop a second faster card.
Please, I hope that is does scare people like you away. I run GNU/Linux because I am free to modify every part of my system. A 'Free' system that relies on proprietary drivers is no longer Free. Users like you help to destroy my freedom.
You can just go back to using Windows or OS X. You are not welcome in our community (what a crazy idea! A community where everyone contributes!).
What if you have a MIPS box? Or a PPC? IA64?
Those systems have PCI, and could use the cards. But they can't anymore because of the x86-only drivers.
What if the driver is unstable on SMP and you are not in the target market for the developer so you can't even make it work with SMP? Hey, you just got fucked out of $NNN for a new card that you can't use unless you want to disable your second CPU.
What if you are doing kernel dev, or just want to test some of the newish stuff like 4kB stacks or regparm argument passing? Out of luck yet again.
What if you are working on a new graphics system, and don't want to use X? Well, with nVidia you are out of luck, and with ATi you are only able to do it if you base your system or DRI (and even then, a lot of times it won't work).
Proprietary software is a festering pile of shit.
No, prison management companies receive a lot of money and then pocket it while treating the prisoners like shit (doing things like keeping medication from sick prisoners).
Remember that punishment has been shown to not rehabilitate people anyway. It merely makes them try harder to not get caught (in most cases).
How the Hell can you suggest that we cut back on food and heating for prisoners? They are still people, and are already being treated as animals (Sodexho...)
If I need to stay up a few extra hours I have found that a pot or two of tea helps greatly. The energy increase is slight and prolonged instead of intense and quickly fading. All of the sugar in the energy drinks makes me sleepy (something with insulin and whatever that makes you tired after a big meal). A nice Yuannin/Darjeeling blend has become my standard for all nighters. As soon as the first woody sip I am reminded of nights past...
The real way to get more energy is to make sure you are taking vitamins (it is difficult to eat a really balanced diet), and do a small amount of exercise. A few pushups and situps in the morning, and a bit of walking during the day do wonders (I have the advantage of going to college, so I can wander around campus between classes). I have found that waking up an hour early, and then spending the hour drinking a pot of tea and breakfast also makes the day better. As tempting as hitting snooze is, putting my shower off and doing things that require no mental or physical effort (put water in the kettle, weigh the leaves, drink a glass of water, toss a bagel in the toaster and preheat the pot right before the kettle boils, set the timer for four minutes and let the tea steep, add cream cheese to the bagel, eat and drink). It is a good way to gradually wake up (the caffeine helps too).
As for why I (*shudder*) at the thought that students are responsible for some of the code in BSD (and also Linux, I shudder at that too), I have seen the code that students write. I've seen this come out of a student's garbled mind (written in C)...
Which students? The stupid ones who shouldn't be in comp sci? Those kind don't actually enjoy coding, and are not the ones who work on GNU/Linux/BSD. The good comp sci students, however, produce good code because they enjoy coding. Remember who was a college student with Linux was first written?
And there are other parts that make it sound like LiveJournal would never be in trouble for this unauthorized access parts. But really, who would bother to post their thoughts and words on a site that has no garauntee of saving them? At any minute, LiveJournal could format its servers and databases and start over with no one able to say anything.
Using a client like Logjam (or emacs lj-update) you can synchronize your journal with an offline copy. The LJ code base is Free Software (well, most of it anyway) so you could always set up your own LJ server. The local entries in logjam are just a directory tree of xml files (with each xml file being a month of entries in a very simple xml format), and it wouldn't be difficult to write a program to translate them into any format you wanted (I could do this fairly quickly using Scheme + SSAX). The only issue left would be external images. I store all of my lj related images on my site in a special dir so I would still have those, and you can now insert images from Scrapbook into your journal so someone could write a ScrapBook backup script, and then you could write a quick perl script to find/replace all of the scrapbook references with whatever host you wanted to copy the images to.
As long as you keep a local backup of your journal (which is easy to do with tools easily available for any system) it would be fairly painless to move them anywhere else. The tasks are simple enough for even a not-so-good programmer to code up into a simple application for others to use.
The amount of pressure needed to floor the gas is only slightly higher than that need to move forward a 1 MPH. Likewise, unless you're driving a huge 40 ton earthmover, braking force to lock the wheels is only slightly greater than the force necessary to gently stop the car.
My 91 Camaro disagrees with you. The pedal has low resistance until around halfway down when the resistance gets progressively larger until it hits the floor. It isn't much, but it is far stiffer than most cars I have driven (it is really strange to drive my car after driving another one for more than a day or two). It isn't so much that you have to stomp, but the force needed to push the pedal past halfway it not insignificant (and yes, I do lubricate the pedal on the schedule my service manual says to).
sssshhhh!
I get a mild buzz from a single bottle of Dogfish Head World Wide Stout. It's made in Delaware, and weighs in at 18% ABV...
American craft beer is good. Bud/Miller/Coors sucks. But so does most commercial beer (fucking InBev...).
I use ffmpeg to transcode episodes of TV shows for my phone (Nokia 7610). They run about 30M per episode, and are surprisingly watchable. The only thing I wished is that I had the Nokia stereo headphones (even though the 7610 only does mono it still outputs to both ears) if only so that I wouldn't have to have one ear open. It helps a lot to kill the four hours between classes I have on Tuesday and Thursday. Having a 512M RS-MMC card comes in handy.
Even if I transcode them at 35kbps (video + audio) they are OK to watch. You can stream that over standard GPRS. At 96kbps they get pretty high quality (keep in mind this is QCIF so the video is really low res, but the screen is too so you don't notice). That can be streamed over EDGE fine.
Your mum's recipies would be a great application of LaTeX. Just hack together a recipe style file, and some macros to make entering recipies easier. Then you could print a nicely formatted book of her recipies without much trouble.