I registered my feelings on bbc's feedback page
on
BBC Links Linux To MyDoom
·
· Score: 3, Insightful
The text of my letter:
I take issue with the broad association between the linux community and the childish and destructive internet worm myDoom made in the article by Stephen Evans titled "Linux cyber-battle turns nasty".
As a long time linux user and advocate I want to point out that while the worm writer may also turn out to be a linux zealot it is important to note that linux users in general condem the destructive impulse that causes someone to write a virus as much as the next person. What causes someone to become enamored with Linux and open source software in general is at it's core the constructive impulse to admire and improve on something that was built by many hands and works extrordinarily well.
It is true that we are almost all disgusted by the shameless and groundless way that SCO is attempting to profit from the sweat of thousands of volunteer programmers. If you look at what SCO is doing you will see that they are claiming as their own and attempting to charge for code that was written in the worlds most open and transparent development process by thousands of individual developers and users who added a bug report here and a line of code there. The community quite rightly has a collective feeling of ownership for the work that we have donated our time to assemble and are indignant to have an insignificant company attempt to steal from us.
We are offended -- but we don't feel the need to express ourselves through vandalism. I know that I speak for the vast majority when I say that I am confident that once SCO stops bluffing and stalling and finally lays down whatever cards they have it will all prove to be a huge farce. The only ones who will have suffered will be those who were taken in by the SCO's executives pathetic stock-pumping ploy and bought the overvalued stock of a failed tech company with nothing to it's name but a pack of ambitious lawyers.
I've done something similar in the past (I didn't bill it as an Xmas present, but I did set up a site for my family) and it doesn't get used. If you want to set it up and maintain it, great. But I'd get them something with a smaller learning curve as well.
I subscribed to the old emusic service for the following reasons.
A. I think that musicians and distributors should get reembursed for music downloads (though NOT at CD comparable prices -- the distribution is much more efficient and economically speaking _should_ be cheaper). The old EMusic represented a great transition from the bad old day of CDs and the coming good days of ubiquitous accepted and legal P2P. I don't use P2P services at the moment. Though once more musicians start releasing music with licenses that allow for free distribution I look forward to using the more efficient P2P distribution mechanism. It will happen. Markets drive towards efficiency and squeeze out the fat. I'm patient.
B. I used the OLD emusic service as an irregular downloader -- I will go for a long time without downloading anything but then get in the mood and download quite a bit. I am sure that my average usage is under the 40 songs per month -- but with my style of use I'd either have to pay a lot or not be able to download what I want. Your new pricing model doesn't work for me. If unused downloads accumulated so that I could use them later I would be interested.
C. I used MP3s as a way to check out obscure Jazz and Folks artists -- who I wouldn't find otherwise (CD stores typically have pretty poor collections). I'll buy a CD from an artist that I've heard and liked -- either live or on MP3s but I rarely spring 16 dollars for an artist I don't know anything about. With the EMusic I could download the album -- knowing that I'd paid fairly for the music and the service -- and if I like it I can keep it and might later on seek out a CD by that artist. If I don't like what I sample I can delete it and move on. Again your new pricing model doesn't work with this usage.
I expect that most other users are in a similar situation and will cancel. I hope that your cash flow position is such that you can withstand the upcoming drought -- at least long enough to come up with (and implement and advertise) a better policy.
To the outgoing management -- good luck, you did a good thing.
Ask yourself what you want the car for? Why are you willing to pay extra for the hybrid?
I'm guessing your answers will be something like:
A) Economics B) The Environment C) Geek Love of Tech
If you are mostly thinking A then you should look at other options. If you are motivated by B or C then the hybrid is for you.
I have had a Honda Insight for about 3 years and I love it. It handles fine in traffic and on the highway. As several posters point out, the economics of getting a hybrid are questionable -- but the mpg savings definitely make the higher service costs and more rapid depreciation less painful. They also lead to a satisfying sense of superiority as prices rise towards 2$/gal.:)
The turning point for me was a sense of frustration about society's general disregard for the environment. I guess I wanted to make a statement -- I beleve it's better to drive a hybrid than to just slap a 'tree hugger' sticker on the back bumper of an SUV.
Also -- I bought my car for commuting (I used to spend 8-10 hours a week in the car). My wife has a slightly larger car which we use whenever we need to haul anything or anyone.
I'm not sure I want this for my PC. However, I can't help thinking that there are occastions when I do want to be able to type with two little joysticks.
Imagine writing email or notes on a PDA or web-pad with two 8 way rockers for your thumbs. Similarly the playstation and xbox controllers we all have would become as capable an input device as a full fledged keyboard.
As a EQOA player who is constantly switching back and forth between my keyboard and my controller I think being able to use my controller as a fast keyboard would rule. Even if it isn't as fast as a traditional keyboard if it is faster than that software keyboard it might take off.
Don't get me wrong. I am not in favor of monopolies, and I'd rather that they would have ruled that the current telcom rules applied to cable.
---- however ----
5 or 10 years ago the average person on the street still didn't -get- the internet. They didn't understand the fundamental difference between a broadcast media and an interactive media.
If this ruling was proposed then it would have made sense to be up in arms because customers wouldn't know what they were missing when providers served them AOL or a lookalike.
Today however, many americans understand. The've all been amazed first hand at the variety and 'fresh-ness' of the internet. Do you think they would let that go away? Don't you think people will vote with their feet against the first such monopolist that started restricting content.
Isn't that the advantage of a free market. The government doesn't have to get it and companies that don't get it will suffer.
I just don't think that this will destroy the spirit of the internet. And with money to be made perhaps the providers will finally run fiber that last mile into everyones living room.
I'm a member of the customer services department at
Etnus,
so I'm interested in helping you get
TotalView
working. You didn't give any information about your system, so it's a little
hard to troubleshoot your problem. But I'll take a stab anyway.
Since you are posting to Slashdot I'd say that the likelihood is good that
you are on top of the latest releases of all things related to Linux, and
that you are comfortable dealing with libraries.
The latest versions of libbfd changed in a way that breaks functionality
that TotalView relies on. If this is indeed what is happening in your case
then the fix is simple. The fix is just to include the library that TotalView
expects to find. Grab libbfd-2.9.5.0.22.so out of
binutils-2.9.5.0.22-6.
You can install it in/usr/lib/libbfd-2.9.5.0.22.so or you can place it
anywhere you want and just update the symbolic link libbfd-2.9.1.0.15.so.0 in/usr/toolworks/totalview.4.1.0-2/linux-x86/shlib/
to point at your libbfd-2.9.5.0.22.so.
Our release package will be changed to include a working libbfd sometime in
the next week.
Please feel free to contact us at
support@etnus.com
about this or any other problem.
Multipole methods can be used in some restricted geometries and for specific problems. However they are not a pancea.
The literature of the field (computational physics) is full of such approximations (fft, bessel function transformations, tree-codes, multipole methods, and dozens of others that I am less familiar with) however there are always trade off issues. Reducing the number of operations or the ammount of communication always comes down to throwing out some of the information. What information can be safely thrown out without jepardizing the validity of the solution is always the meat of the question.
So yes, in some cases the latency can be beaten down with multipole expansions. I simply point out that highly coupled problems exist, they are interesting, and they do not all have 'convenient' geometries.
In the end it comes down to something that is really interesting about parallel computing in general. Parallel computers aren't really general purpose beasts, there is a huge range of architectures each with different characteristics. Similarly there are huge range of problems and algorithms and for each class of problem a different kind of computer will be most effective.
Specifically wide area distributed computing will probably never be useful for evolving forward highly coupled dynamical systems because of latency. These systems probably need dedicated machines with stripped down network protocols or even a hardware message passing or shared memory architecture.
However distributed computing would be marvelous for exploring huge regions of parameter space with 'smaller' problems, if the problem can fit on one machine (even if it takes weeks to solve) then you can try out millions of different initial conditions and really map out the behavior of the system. An example from my realm of such a problem is modeling gravitational lenses. The methodology is to solve a simple problem (raytracing with a general relativistic lens) for a range of parameters for the lens galaxy and its surroundings, then find the model which most accrately fits the image. This has of course been done in parallel for a long time, at one point using floppy disks and a lab full of PC's (sneakernet protocol). Of course that wasn't millions of machines.
Nonetheless you are right it is an issue. What would be ideal would be an open source toolbox; reviewed and trusted (in the cryptographic sense). Ideally you would want it to be able to provide the compute services to the distributed project but would wall the process away from the rest of the OS and the users data. Of course an open source toolbox would be vulerable to users looking at what the distributed program was doing. So companies wanting to do propritary work on the system would have to break the problem up into indecyperable pieces.
Actually given that the machine is always going to be fundamentally under the users control there is really no way for the company to avoid a determined user finding out what his computer is doing. Even if it is transmitted securely the computer can hardly hide the operations its performing or the data that it is performing them on from itself.
I think that these schemes should only be used for non-sensitive datasets and should always be run with redundancy as the article suggests (to catch computational errors/tampering).
from the prospective of a physicist, unfortunately the answer is 'most of them'.
The sort of problems that applied physicists (in my case astrophysicists) want to use supercomputers for are those that involve the 'largest systems'. In this context I'm referring to a system as a set of interdependent parts. Take supernovae calculations for instance. The 'parts' in the calculations are segments of the star, each of which has a state that is dependent upon what is going on in other 'parts' of the star. A specific grid cell needs to know about the pressure forces that it is feeling from all of the other cells right next to it. But it also needs to know about the gravitational field generated by parts that might be on the other side of the star. Radiation and particles like neutrinos also have a nasty habit of traveling quite a ways away from their source and influencing things 'at a distance'.
The problem is that everything is dynamic. You have to find a solution for what is going on in each part then all of the parts have to communicate with each other (latency, lots and lots of it) then you have to repeat this process many many times (tens of thousands of itterations may not be enough to capture both the fast and the slow processes that the physicist is interested in understanding).
This is a really common example. That's why there is so much interest not only in high speed interconnects, but low latency interconnects among the beowulf community.
Actually there are pretty reasonable arguements against that possibility. The strongest of which is cosmological. If you believe cosmology the relative abundance of the light elements (hydrogen, helium, and lithium) would be thrown all out of whack if the universe has more baryons (stars, dust, moons, gas and such) than about a third of the mass needed to make the universe flat. Basically this has to do with the fact that during the first few moments (of the universe) nuclear reactions are going on all of the time turning hydrogen into helium and back again. The forward and backwards reactions are density dependent and go on for as long as the universe is hot enough to sustain the reaction. So the relative abundances give you a measure of the density of the baryons in the universe at the moment that the universe cooled enough to stop the reaction.
The universe appears to be flat, from redshift surveys and the ripples in the cosmic microwave background. So since we know that baryons can't do more than a third of that we are forced to postulate something wierd to account for at least the other two thirds.
The most likely value of the baryon fraction is actually around 12 percent and the rest is split between dark matter and something even wierder called 'vacuume energy of space' (or 'lambda').
Actually most of the mass in a typical galaxy is neither stars nor dust, but rather some as of yet unknown form of matter called 'dark matter'.
As far as we know the 'gravity only' type of calculation that the Grape boards perform is sufficient to describe the motion of this matter.
However, there is indeed great interest in performing hydrodynamical simulations of galaxies, mostly because then we can attempt to calculate where and how the stars are forming in the galaxy. Dealing with the gas expicitly also allows us to follow things like shock fronts in the gas and to attempt to calculate the thermal properties of the gas. Of course this is all rather complicated stuff so we have to make gross approximations. And remember, even with that massive grape board if you describe a galaxy with a million particles, they are still going to each be representing at least 10 to 100 thousand solar masses. We are still a long way off from being able to describe the milky way on a 'star by star' basis.
Two really good URL's for people who are interested in reading some of the technical details of this stuff are the web pages of my advisor Matthias Steinmetz and one of the fathers of modern galaxy simulations Josh Barnes.
Note that Matthias's simulations (check out the pictures and movies) are all done with a high end workstation and a handfull of Grape 3 boards.
It's my understanding that the highest redshift (therefore oldest/earliest/farthest away) objects other than the CMB are galaxies. I think that there are confirmed redshifts for galaxies at around six. I can't find a good reference for that figure. What I was able to find in a few minutes searching was a catalog of galaxies from the hubble deep field north. In this one paper Lanzetta et. al. claim
"We have identified nearly 3000 faint galaxies, of which nearly 1000 galaxies are of redshift z > 2 and more than 50 galaxies are of redshift z > 5 (ranging up to and beyond z = 10)."
Now, I should caution you that these are photometric redshifts, somewhat more speculative than those derived from matching spectral lines (as was done with the quasar atz=5.5). But in principle with large enough telescopes we can go back and do the spectroscopy and verify these redshifts, they won't all be correct but most of them will.
Another interesting possibility that you should look at if you are interested in having a clever answer to the question 'what's the furtherst thing in the universe' are gamma ray bursts. Though redshifts for these are hard to get it is possible to make speculative arguments about their redshift distribution based on the idea that some of them may appear to 'last longer' due to cosmological time dialation.
Finally, as several others have already pointed out the Cosmic Microwave Background estimated to be at z of about 1500, is about as far away as we are going be able to see. Farther off, you are looking back into the universe when it was so hot and dense that it was 'opaque'. The CMB represents the point in the evolution of the universe when things cooled enough for neutral atoms to form. It turns out that electrons running around without a proton make it really hard for photons to get anywhere in a straight line. At z=1500 those free electrons got used up to make neutral hydrogen and the universe suddenly became 'transparent'.
Incidentally, the universe had to become transparent for the gas to ever cool and form galaxies, stars, planets, and people. In this way and many others the cosmic microwave background represents the beginning of all of the structure that we see around us.
Slashdot Mirror
The text of my letter:
I take issue with the broad association between the linux community and the childish and destructive internet worm myDoom made in the article by Stephen Evans titled "Linux cyber-battle turns nasty".
As a long time linux user and advocate I want to point out that while the worm writer may also turn out to be a linux zealot it is important to note that linux users in general condem the destructive impulse that causes someone to write a virus as much as the next person. What causes someone to become enamored with Linux and open source software in general is at it's core the constructive impulse to admire and improve on something that was built by many hands and works extrordinarily well.
It is true that we are almost all disgusted by the shameless and groundless way that SCO is attempting to profit from the sweat of thousands of volunteer programmers. If you look at what SCO is doing you will see that they are claiming as their own and attempting to charge for code that was written in the worlds most open and transparent development process by thousands of individual developers and users who added a bug report here and a line of code there. The community quite rightly has a collective feeling of ownership for the work that we have donated our time to assemble and are indignant to have an insignificant company attempt to steal from us.
We are offended -- but we don't feel the need to express ourselves through vandalism. I know that I speak for the vast majority when I say that I am confident that once SCO stops bluffing and stalling and finally lays down whatever cards they have it will all prove to be a huge farce. The only ones who will have suffered will be those who were taken in by the SCO's executives pathetic stock-pumping ploy and bought the overvalued stock of a failed tech company with nothing to it's name but a pack of ambitious lawyers.
I've done something similar in the past (I didn't bill it as an Xmas present, but I did set up a site for my family) and it doesn't get used. If you want to set it up and maintain it, great. But I'd get them something with a smaller learning curve as well.
Just my 2c.
I subscribed to the old emusic service for the following reasons.
A. I think that musicians and distributors should get reembursed for music downloads (though NOT at CD comparable prices -- the distribution is much more efficient and economically speaking _should_ be cheaper). The old EMusic represented a great transition from the bad old day of CDs and the coming good days of ubiquitous accepted and legal P2P. I don't use P2P services at the moment. Though once more musicians start releasing music with licenses that allow for free distribution I look forward to using the more efficient P2P distribution mechanism. It will happen. Markets drive towards efficiency and squeeze out the fat. I'm patient.
B. I used the OLD emusic service as an irregular downloader -- I will go for a long time without downloading anything but then get in the mood and download quite a bit. I am sure that my average usage is under the 40 songs per month -- but with my style of use I'd either have to pay a lot or not be able to download what I want. Your new pricing model doesn't work for me. If unused downloads accumulated so that I could use them later I would be interested.
C. I used MP3s as a way to check out obscure Jazz and Folks artists -- who I wouldn't find otherwise (CD stores typically have pretty poor collections). I'll buy a CD from an artist that I've heard and liked -- either live or on MP3s but I rarely spring 16 dollars for an artist I don't know anything about. With the EMusic I could download the album -- knowing that I'd paid fairly for the music and the service -- and if I like it I can keep it and might later on seek out a CD by that artist. If I don't like what I sample I can delete it and move on. Again your new pricing model doesn't work with this usage.
I expect that most other users are in a similar situation and will cancel. I hope that your cash flow position is such that you can withstand the upcoming drought -- at least long enough to come up with (and implement and advertise) a better policy.
To the outgoing management -- good luck, you did a good thing.
CANCEL MY SUBSCRIPTION EFFECTIVE IMMEDIATELY!
Ask yourself what you want the car for? Why are you willing to pay extra for the hybrid?
:)
I'm guessing your answers will be something like:
A) Economics
B) The Environment
C) Geek Love of Tech
If you are mostly thinking A then you should look at other options. If you are motivated by B or C then the hybrid is for you.
I have had a Honda Insight for about 3 years and I love it. It handles fine in traffic and on the highway. As several posters point out, the economics of getting a hybrid are questionable -- but the mpg savings definitely make the higher service costs and more rapid depreciation less painful. They also lead to a satisfying sense of superiority as prices rise towards 2$/gal.
The turning point for me was a sense of frustration about society's general disregard for the environment. I guess I wanted to make a statement -- I beleve it's better to drive a hybrid than to just slap a 'tree hugger' sticker on the back bumper of an SUV.
Also -- I bought my car for commuting (I used to spend 8-10 hours a week in the car). My wife has a slightly larger car which we use whenever we need to haul anything or anyone.
Best of luck in your decision!
I'm not sure I want this for my PC. However, I can't help thinking that there are occastions when I do want to be able to type with two little joysticks.
Imagine writing email or notes on a PDA or web-pad with two 8 way rockers for your thumbs. Similarly the playstation and xbox controllers we all have would become as capable an input device as a full fledged keyboard.
As a EQOA player who is constantly switching back and forth between my keyboard and my controller I think being able to use my controller as a fast keyboard would rule. Even if it isn't as fast as a traditional keyboard if it is faster than that software keyboard it might take off.
What do you think?
Don't get me wrong. I am not in favor of monopolies, and I'd rather that they would have ruled that the current telcom rules applied to cable.
---- however ----
5 or 10 years ago the average person on the street still didn't -get- the internet. They didn't understand the fundamental difference between a broadcast media and an interactive media.
If this ruling was proposed then it would have made sense to be up in arms because customers wouldn't know what they were missing when providers served them AOL or a lookalike.
Today however, many americans understand. The've all been amazed first hand at the variety and 'fresh-ness' of the internet. Do you think they would let that go away? Don't you think people will vote with their feet against the first such monopolist that started restricting content.
Isn't that the advantage of a free market. The government doesn't have to get it and companies that don't get it will suffer.
I just don't think that this will destroy the spirit of the internet. And with money to be made perhaps the providers will finally run fiber that last mile into everyones living room.
Just a thought.
I'm a member of the customer services department at Etnus, so I'm interested in helping you get TotalView working. You didn't give any information about your system, so it's a little hard to troubleshoot your problem. But I'll take a stab anyway.
Since you are posting to Slashdot I'd say that the likelihood is good that you are on top of the latest releases of all things related to Linux, and that you are comfortable dealing with libraries.
The latest versions of libbfd changed in a way that breaks functionality that TotalView relies on. If this is indeed what is happening in your case then the fix is simple. The fix is just to include the library that TotalView expects to find. Grab libbfd-2.9.5.0.22.so out of binutils-2.9.5.0.22-6.
You can install it in /usr/lib/libbfd-2.9.5.0.22.so or you can place it
anywhere you want and just update the symbolic link libbfd-2.9.1.0.15.so.0 in /usr/toolworks/totalview.4.1.0-2/linux-x86/shlib/
to point at your libbfd-2.9.5.0.22.so.
Our release package will be changed to include a working libbfd sometime in the next week.
Please feel free to contact us at support@etnus.com about this or any other problem.
binutils and libbfd are licensed under the GPL.
Cheers,
Chris Gottbrath, Etnus
The literature of the field (computational physics) is full of such approximations (fft, bessel function transformations, tree-codes, multipole methods, and dozens of others that I am less familiar with) however there are always trade off issues. Reducing the number of operations or the ammount of communication always comes down to throwing out some of the information. What information can be safely thrown out without jepardizing the validity of the solution is always the meat of the question.
So yes, in some cases the latency can be beaten down with multipole expansions. I simply point out that highly coupled problems exist, they are interesting, and they do not all have 'convenient' geometries.
In the end it comes down to something that is really interesting about parallel computing in general. Parallel computers aren't really general purpose beasts, there is a huge range of architectures each with different characteristics. Similarly there are huge range of problems and algorithms and for each class of problem a different kind of computer will be most effective.
Specifically wide area distributed computing will probably never be useful for evolving forward highly coupled dynamical systems because of latency. These systems probably need dedicated machines with stripped down network protocols or even a hardware message passing or shared memory architecture.
However distributed computing would be marvelous for exploring huge regions of parameter space with 'smaller' problems, if the problem can fit on one machine (even if it takes weeks to solve) then you can try out millions of different initial conditions and really map out the behavior of the system. An example from my realm of such a problem is modeling gravitational lenses. The methodology is to solve a simple problem (raytracing with a general relativistic lens) for a range of parameters for the lens galaxy and its surroundings, then find the model which most accrately fits the image. This has of course been done in parallel for a long time, at one point using floppy disks and a lab full of PC's (sneakernet protocol). Of course that wasn't millions of machines.
They do mention user security.
Nonetheless you are right it is an issue. What would be ideal would be an open source toolbox; reviewed and trusted (in the cryptographic sense). Ideally you would want it to be able to provide the compute services to the distributed project but would wall the process away from the rest of the OS and the users data. Of course an open source toolbox would be vulerable to users looking at what the distributed program was doing. So companies wanting to do propritary work on the system would have to break the problem up into indecyperable pieces.
Actually given that the machine is always going to be fundamentally under the users control there is really no way for the company to avoid a determined user finding out what his computer is doing. Even if it is transmitted securely the computer can hardly hide the operations its performing or the data that it is performing them on from itself.
I think that these schemes should only be used for non-sensitive datasets and should always be run with redundancy as the article suggests (to catch computational errors/tampering).
The sort of problems that applied physicists (in my case astrophysicists) want to use supercomputers for are those that involve the 'largest systems'. In this context I'm referring to a system as a set of interdependent parts. Take supernovae calculations for instance. The 'parts' in the calculations are segments of the star, each of which has a state that is dependent upon what is going on in other 'parts' of the star. A specific grid cell needs to know about the pressure forces that it is feeling from all of the other cells right next to it. But it also needs to know about the gravitational field generated by parts that might be on the other side of the star. Radiation and particles like neutrinos also have a nasty habit of traveling quite a ways away from their source and influencing things 'at a distance'.
The problem is that everything is dynamic. You have to find a solution for what is going on in each part then all of the parts have to communicate with each other (latency, lots and lots of it) then you have to repeat this process many many times (tens of thousands of itterations may not be enough to capture both the fast and the slow processes that the physicist is interested in understanding).
This is a really common example. That's why there is so much interest not only in high speed interconnects, but low latency interconnects among the beowulf community.
Actually there are pretty reasonable arguements against that possibility. The strongest of which is cosmological. If you believe cosmology the relative abundance of the light elements (hydrogen, helium, and lithium) would be thrown all out of whack if the universe has more baryons (stars, dust, moons, gas and such) than about a third of the mass needed to make the universe flat. Basically this has to do with the fact that during the first few moments (of the universe) nuclear reactions are going on all of the time turning hydrogen into helium and back again. The forward and backwards reactions are density dependent and go on for as long as the universe is hot enough to sustain the reaction. So the relative abundances give you a measure of the density of the baryons in the universe at the moment that the universe cooled enough to stop the reaction.
The universe appears to be flat, from redshift surveys and the ripples in the cosmic microwave background. So since we know that baryons can't do more than a third of that we are forced to postulate something wierd to account for at least the other two thirds.
The most likely value of the baryon fraction is actually around 12 percent and the rest is split between dark matter and something even wierder called 'vacuume energy of space' (or 'lambda').
Hope that made some sense.
chris
it's your universe, get used to it
As far as we know the 'gravity only' type of calculation that the Grape boards perform is sufficient to describe the motion of this matter.
However, there is indeed great interest in performing hydrodynamical simulations of galaxies, mostly because then we can attempt to calculate where and how the stars are forming in the galaxy. Dealing with the gas expicitly also allows us to follow things like shock fronts in the gas and to attempt to calculate the thermal properties of the gas. Of course this is all rather complicated stuff so we have to make gross approximations. And remember, even with that massive grape board if you describe a galaxy with a million particles, they are still going to each be representing at least 10 to 100 thousand solar masses. We are still a long way off from being able to describe the milky way on a 'star by star' basis.
Two really good URL's for people who are interested in reading some of the technical details of this stuff are the web pages of my advisor Matthias Steinmetz and one of the fathers of modern galaxy simulations Josh Barnes.
Note that Matthias's simulations (check out the pictures and movies) are all done with a high end workstation and a handfull of Grape 3 boards.
Cheers
"We have identified nearly 3000 faint galaxies, of which nearly 1000 galaxies are of redshift z > 2 and more than 50 galaxies are of redshift z > 5 (ranging up to and beyond z = 10)."
Now, I should caution you that these are photometric redshifts, somewhat more speculative than those derived from matching spectral lines (as was done with the quasar atz=5.5). But in principle with large enough telescopes we can go back and do the spectroscopy and verify these redshifts, they won't all be correct but most of them will.
Another interesting possibility that you should look at if you are interested in having a clever answer to the question 'what's the furtherst thing in the universe' are gamma ray bursts. Though redshifts for these are hard to get it is possible to make speculative arguments about their redshift distribution based on the idea that some of them may appear to 'last longer' due to cosmological time dialation.
Finally, as several others have already pointed out the Cosmic Microwave Background estimated to be at z of about 1500, is about as far away as we are going be able to see. Farther off, you are looking back into the universe when it was so hot and dense that it was 'opaque'. The CMB represents the point in the evolution of the universe when things cooled enough for neutral atoms to form. It turns out that electrons running around without a proton make it really hard for photons to get anywhere in a straight line. At z=1500 those free electrons got used up to make neutral hydrogen and the universe suddenly became 'transparent'.
Incidentally, the universe had to become transparent for the gas to ever cool and form galaxies, stars, planets, and people. In this way and many others the cosmic microwave background represents the beginning of all of the structure that we see around us.
I just keep wishing that people wrote straightforward html so that I could change the font size myself.
Anyway the tech looks pretty snazy. I want one.