In short: it works just fine, and sits (moderately quietly) in the corner, doing its thing. However, the processor is not really fast enough for desktop, so installing and getting everything working was a little frustrating. Also, that chip doesn't have automatic throttling support in Linux so I have a bit of a hack (see link). And the original fan was small and noisy, so I replaced it with a 120mm one.
My electricity bill dropped by about 1/3 since my main machine, which I use during the day for development work, is off for over half the time.
it's not clear to me (sorry if i've missed some info somewhere) whether this is paid work or not.
if it's a personal project, then one option is to take a break. i've been writing software for way too long (20 years?) and in that time have nearly always programmed in my free time, but there was a period of about a year when i felt burn out and spent my free time building electronics and learning bass guitar.
another option - one that is possible if it's a personal project and may also be possible (perhaps with some negotiation) if it's paid - is so switch focus. sometimes, for example, writing documentation can give you a new view on a project that gets you back working. i am currently working on a parser (called lepl) and whenever i work on documentation i end up with a whole pile of new ideas.
related to that - dog fooding. can you use the software you are developing in some way? that can motivate you to add a feature that would really help you out. similarly, getting a release out. particularly if it's open source (or a free beta?) - feedback from other people is motivating and helps highlight the most important features (which might not be what you were expecting).
at the same time - trust yourself. my guess is that you've not been programming that long. these things always come in cycles. once you've been through a few it's much easier to just take it easy, knowing that the drive will come back. i'm lucky in that i work from home - there are days when i do almost nothing but eat, sleep and surf the web (and there are many more days when i work my ass off - the idea is that there's give and take, it's a two sided deal....)
finally, look after yourself. look at what you're eating, how well you're sleeping, whether you're getting enough exercise. perhaps take a personal day and do something interesting.
No really exciting differences, but the text without Slashdot's edits (which removed links, sub-headings and a paragraph that explained what I was doing) can be found here.
Fortress, the language being developed by a bunch of people led by Guy Steele, was funded as part of the HPCS effort. This means that DARPA is going with IBM or Cray's language (X10 for IBM, Chapel from Cray). According to a press release quoted at http://www.eweek.com/article2/0,1895,2063043,00.as p (but not available at http://www.sun.com/smi/Press/sunflash/index.xml) the work will continue, but how likely is it to succeed?
the competition in astronomy is fierce. there's a fixed amount of money and a pile of good projects. there's a big peer-review process that evaluates possible projects and gives priorities. then the nsf goes round looking for dead wood it can hack away so that there's money for the best projects. no-one is complacent - i work at ctio and everyone there was assuming that they were going to lose their jobs. and because lsst won't really kick in for a few years, we may still be laid off before then (even though we're all working like crazy on related projects). this isn't a bunch of "has beens" making life easy for themselves - it's a vicious, competitive world where only projects that really stand a good chance of changing astronomy make it.
second, the technology choice:
if you are talking about synthetic apertures (like radio telescopes) then no - you cannot link optical telescopes together state-wide. you can control them in parallel, sure, but you cannot combine the data in the same way as radio telescopes. it's way beyond our technical ability. so if there is no synthetic aperture, what's the advantage in spreading them around? especially when world class telescope sites with existing support are very rare. it makes most sense to put one telescope on the top of a mountain in a chilean desert.
and don't think you can re-use any old telescope. the structural engineering of this thing is going to be brutal - to optimize throughput the slews (moving to a new position on the sky) are going to be way faster than anything currently out there. that's one reason the site decision had to be made early - they need to know what they're building this on just to control the vibration levels!
there is a competing project, called pan-stars, which has a group of co-located telescopes. the advantage of that approach is largely political - you can build one cheaply and then look for more funding. but if you do the maths - and this is well understood engineering/optics/statistics, the answer is clear - the lsst solution comes out on top.
oh, and it's not old news either; the press conference anouncing that this was going to chile was held in the room next to my office a few days ago.
i understand (i work at the ctio observatory as a programmer and obviously everyone there is very excited by this news) that the lsst will require an impressive amount of computing power, but that it's not impractical - it would be possible today, although expensive, and the hope is that in a few years the hardware will be much more reasonably priced. since processing a lot of images is pretty easy to do in parallel it's likely that the hardware will be some kind of cluster rather than a traditional (old school) supercomputer (we already have parallel pipelines for processing astronomical data - see work by frank valdes at noao, for example). obviously there's a trade-off between how much processing power you do and how much bandwidth you have (since processed data is more compact) - i believe that determines the location of the computers (i'm unsure how much will be up on the mountain and how much "downtown" at the observatory's offices in la serena). as for internet connections - chile has two major commerical providers who can handle the capacity required.
personally, i'm currently part of the team working on an archive for the noao's existing telescopes (the noao in conjunction with the ncsa) and we're hoping that will help provide a basis for the data archive that the lsst will use. what's particularly cool about the lsst (at least last i heard) is that the data are public almost immediately - typically that's not the case for telescopes, where the astronomer doing the observations gets a year or so to use the data for themselves first - and that makes the archive critical. so it's a whole new paradigm - astronomers will "observe" by mining the database.
sorry i don't have more exact answers - i've sat through presentations on this to the point where my eyes glaze over at the numbers. i'm going to see if i can dig anything up that will bemore helpful...
The most interesting link here is behind a pay-wall. Do the editors bother to follow the link in articles? Do they just assume we all have ACM access? Come on, this place used to be a bit better than this, didn;t it?
incidentally, chile recently had elections for a new president (and elected, as you may have heard, a single parent socialist woman).
anyway, that was the first election i've experienced here in chile (i'm english, but have lived here for several years, and have a chilean partner). i was *very* impressed with the way the elections were run here. the counting is done in public, with every vote being read out and held up to public view, by randomly selected members of the public.
it was jaw-dropping to watch tv and see ordinary chileans sitting around a table counting votes with pencil and paper, and the public crowding round watching. i have never seen such a clear, clean, openly democratic process.
at the same time, the results were available the same day the election closed. by early evening the outcome was clear. and this is with simple paper ballots counted by hand.
finally, and even more amazingly, next time there are elections for president, i will be able to vote. all you need to do to qualify is live here for five years (i admit i've not checked the details, but that is what i understand). on the other hand, chileans abroad cannot vote.
oh, and voting is compulsory, but only if you register. registering is optional. which seems like a good compromise to me.
(generally i seem to complain about this bloody country, but they deserve a huge amount of credit for the way they run elections. i wish it were this clear and involved as much mass-participation in the uk).
it depends very much on the job, for two reasons. first, you probably aren't going to be writing code all the time. you might spend a lot of time doing design work, requirements gathering, use cases, yadda yadda, or testing, or just sitting in meetings. second, work loads vary by huge amounts, depending on the company, position, etc.
my previous job ended up with 6 day weeks, working over 12 hours a day (i would guess, on average - at times we were there for 24 hours or more). most of that was writing code (with no design, etc etc). no way did i write code in my "spare time" then (and, even when i left, after having had a minor breakdown, for months i had no desire to program).
live and learn - my next (current job) i negotiated as shift working, with longish hours for 8 days, then 6 off. on average, i work average hours, but with (1) less room to work extra hours and (2) lots of time between shifts for my own projects. the employer cares more about design/process and it also pays more than twice as much (but that's due to rather odd circumstances - i live in s. america).
so, in the end, it comes down to what you make of it. you can get sunk into a project (like i was) where you're "responsible for saving everyone's jobs" (the world didn't end when i left, though), and it can ruin your life. or you can take a bit more control and have plenty of time (and energy) left for what you want to do.
(my current project is a little programming language, a bit like forth).
other people have commented that you get to see more than that due to rotation of the earth during the night, and movement of th eearth around the sun during the year.
however, there's a more fundamental reason why this is largely unimportant - the universe doesn't have a special direction, it's pretty much the same everywhere. so while you might not be able to see a certain object from a certain telescope, you can see another one pretty much like it.
there are exceptions, of course. if you're looking at objects in our galaxy then you may need to use a certain telescope, because the position of the galaxy relative to the earth is pretty much fixed. so for "nearby" objects it may be important. also, at the extreme opposite, observations of large scale fluctations in the very early universe (effectively observing *very* distance objects) may require all-sky observations.
but for many interesting objects - other galaxies, quasars, radio galaxies, etc there's no real loss to being restricted to one particular direction.
Google somehow (probably a though a CSS hack) manages to substitute a 1x1 white pixel.GIF file for the page if you try to print it or copy its location.
the page image itself is set as the background image - if you look at the page source (via the "view" menu in the toolbar, search for the embedded CSS fragment "background-image:url") you can get the URL, load it explicitly, and get direct access.
the telescope/instrument control systems are/will not be linux, as far as i know, but the data will be archived and reduced on linux (some reduction might also be on suns and max osx).
i work at ctio, writing iraf software for noao - iraf is multi-platform, but we develop on linux (currently red hat, about to move to fedora, although i also have it running in debian on my laptop - well, i did until yesterday, when i messed up a kernel recompile/install and lost linux completely (it's an x31, so i need to do a network boot to get it back etc etc):o)
hmmm. i think your c algorithm is O(n^2) too, but only in the limit of large "count" - then the inner loop is O(n) (in practice, count is not that large, and that inner loop has a very small constant, so that loop is proabably not visible in any timing test with reasonable numbers). but i might not have understood what you're doing, and i'm not 100% sure the inner loop is O(n) even then, because the step size is increasing.
i think it's more likely you're seeing a *huge* difference in constant factors. that's very optimized (i'm impressed) c code - of course that also implies it has a whole pile of restrictions (fixed upper limits on values etc).
i doubt any functional solution is going to give you comparable speed to that code - certainly not at the moment. but you're kind-of comparing apples + oranges (it's a bit sneaky that the c code requires an upper bound from the start).
i agree (that write once variables are a feature, not a bug), but it took me a long time to see that and i was trying to relate to how people who haven't used functional programming before might feel. i don't need to tell you it's good, but i think i do need to explain to people who haven't used a functional language that it is ok to find it odd at first (i certainly did).
i'm curious - what's the c version look like? (the haskell version looks ok to me - it seems to be about n^2 according to timings, and i think that's what you'd expect because the number of tests goes up more-or-less as n).
thanks. another good book (that i should have mentioned - i forgot about it when i was writing that, but just used it a moment ago) is cousineau + mauny's "the functional approach to programming". it's like an ml/ocaml version of the (relatively) famous "structure and interpretation of computer programs".
Here (Santiago, Chile), I get an externally visible address for my cable modem, but it's via DHCP, so can change (in practice, it does so very rarely, so I use it as webmail server; my public site is with a colo in the USA). I can't remember what the contract says, but I wouldn't be surprised if I was only supposed to have one machine connected (that was the case in the UK, when I was there). However, nobody has complained about my using NAT (I have a Linux box as firewall with two internal subnets - one has the Linux box with a Windows machine, the other is connected to a wireless hub; the wireless subnet is filtered by MAC with *no* direct local access to the firewall machine).
Hi, Do you have a reference? (I just started work for NOAO in their data products group, so I'm curious what you're doing). Going to ADASS? Cheers, Andrew
After something similar (but due to bad management rather than bad economy), I resigned and took a months holiday. At the end of the month I returned, negotiated better working conditions (double time for overtime and working from home) and started again.
However the intense period before - when we were working to meet a deadline - was too much (that holiday was to mentally recover - I did little except sleep and occasionally cry/shout/hit walls). I no longer care much about the company and don't enjoy working there (despite the better conditions).
So I'm looking for another job.
Conclusions, then:
- If you're critical to the business you can dictate terms (reasonably, of course)
- Don't leave it too long before doing something, or the harm may already be done.
Slashdot Mirror
just heard the news. not been here in years. peace, roblimo.
I guess I am probably too late to be noticed, but for what it's worth I did exactly this. My notes are here - http://www.acooke.org/cute/SystemRefa0.html
In short: it works just fine, and sits (moderately quietly) in the corner, doing its thing. However, the processor is not really fast enough for desktop, so installing and getting everything working was a little frustrating. Also, that chip doesn't have automatic throttling support in Linux so I have a bit of a hack (see link). And the original fan was small and noisy, so I replaced it with a 120mm one.
My electricity bill dropped by about 1/3 since my main machine, which I use during the day for development work, is off for over half the time.
it's not clear to me (sorry if i've missed some info somewhere) whether this is paid work or not.
if it's a personal project, then one option is to take a break. i've been writing software for way too long (20 years?) and in that time have nearly always programmed in my free time, but there was a period of about a year when i felt burn out and spent my free time building electronics and learning bass guitar.
another option - one that is possible if it's a personal project and may also be possible (perhaps with some negotiation) if it's paid - is so switch focus. sometimes, for example, writing documentation can give you a new view on a project that gets you back working. i am currently working on a parser (called lepl) and whenever i work on documentation i end up with a whole pile of new ideas.
related to that - dog fooding. can you use the software you are developing in some way? that can motivate you to add a feature that would really help you out. similarly, getting a release out. particularly if it's open source (or a free beta?) - feedback from other people is motivating and helps highlight the most important features (which might not be what you were expecting).
at the same time - trust yourself. my guess is that you've not been programming that long. these things always come in cycles. once you've been through a few it's much easier to just take it easy, knowing that the drive will come back. i'm lucky in that i work from home - there are days when i do almost nothing but eat, sleep and surf the web (and there are many more days when i work my ass off - the idea is that there's give and take, it's a two sided deal....)
finally, look after yourself. look at what you're eating, how well you're sleeping, whether you're getting enough exercise. perhaps take a personal day and do something interesting.
No really exciting differences, but the text without Slashdot's edits (which removed links, sub-headings and a paragraph that explained what I was doing) can be found here.
Fortress, the language being developed by a bunch of people led by Guy Steele, was funded as part of the HPCS effort. This means that DARPA is going with IBM or Cray's language (X10 for IBM, Chapel from Cray). According to a press release quoted at http://www.eweek.com/article2/0,1895,2063043,00.as p (but not available at http://www.sun.com/smi/Press/sunflash/index.xml) the work will continue, but how likely is it to succeed?
l Slides9Jun2006.pdf - I thought it was pretty impressive.
Guy Steele gave an excellent talk at OOPSLA on Fortress - the slides are at http://research.sun.com/projects/plrg/PLDITutoria
The groups's site is at http://research.sun.com/projects/plrg/
first, on the whole political thing:
the competition in astronomy is fierce. there's a fixed amount of money and a pile of good projects. there's a big peer-review process that evaluates possible projects and gives priorities. then the nsf goes round looking for dead wood it can hack away so that there's money for the best projects. no-one is complacent - i work at ctio and everyone there was assuming that they were going to lose their jobs. and because lsst won't really kick in for a few years, we may still be laid off before then (even though we're all working like crazy on related projects). this isn't a bunch of "has beens" making life easy for themselves - it's a vicious, competitive world where only projects that really stand a good chance of changing astronomy make it.
second, the technology choice:
if you are talking about synthetic apertures (like radio telescopes) then no - you cannot link optical telescopes together state-wide. you can control them in parallel, sure, but you cannot combine the data in the same way as radio telescopes. it's way beyond our technical ability. so if there is no synthetic aperture, what's the advantage in spreading them around? especially when world class telescope sites with existing support are very rare. it makes most sense to put one telescope on the top of a mountain in a chilean desert.
and don't think you can re-use any old telescope. the structural engineering of this thing is going to be brutal - to optimize throughput the slews (moving to a new position on the sky) are going to be way faster than anything currently out there. that's one reason the site decision had to be made early - they need to know what they're building this on just to control the vibration levels!
there is a competing project, called pan-stars, which has a group of co-located telescopes. the advantage of that approach is largely political - you can build one cheaply and then look for more funding. but if you do the maths - and this is well understood engineering/optics/statistics, the answer is clear - the lsst solution comes out on top.
oh, and it's not old news either; the press conference anouncing that this was going to chile was held in the room next to my office a few days ago.
here's a background paper on the "data challenge" - http://www.lsst.org/Project/docs/data-challenge.pd f
i understand (i work at the ctio observatory as a programmer and obviously everyone there is very excited by this news) that the lsst will require an impressive amount of computing power, but that it's not impractical - it would be possible today, although expensive, and the hope is that in a few years the hardware will be much more reasonably priced. since processing a lot of images is pretty easy to do in parallel it's likely that the hardware will be some kind of cluster rather than a traditional (old school) supercomputer (we already have parallel pipelines for processing astronomical data - see work by frank valdes at noao, for example). obviously there's a trade-off between how much processing power you do and how much bandwidth you have (since processed data is more compact) - i believe that determines the location of the computers (i'm unsure how much will be up on the mountain and how much "downtown" at the observatory's offices in la serena). as for internet connections - chile has two major commerical providers who can handle the capacity required.
personally, i'm currently part of the team working on an archive for the noao's existing telescopes (the noao in conjunction with the ncsa) and we're hoping that will help provide a basis for the data archive that the lsst will use. what's particularly cool about the lsst (at least last i heard) is that the data are public almost immediately - typically that's not the case for telescopes, where the astronomer doing the observations gets a year or so to use the data for themselves first - and that makes the archive critical. so it's a whole new paradigm - astronomers will "observe" by mining the database.
sorry i don't have more exact answers - i've sat through presentations on this to the point where my eyes glaze over at the numbers. i'm going to see if i can dig anything up that will bemore helpful...
The most interesting link here is behind a pay-wall. Do the editors bother to follow the link in articles? Do they just assume we all have ACM access? Come on, this place used to be a bit better than this, didn;t it?
incidentally, chile recently had elections for a new president (and elected, as you may have heard, a single parent socialist woman).
anyway, that was the first election i've experienced here in chile (i'm english, but have lived here for several years, and have a chilean partner). i was *very* impressed with the way the elections were run here. the counting is done in public, with every vote being read out and held up to public view, by randomly selected members of the public.
it was jaw-dropping to watch tv and see ordinary chileans sitting around a table counting votes with pencil and paper, and the public crowding round watching. i have never seen such a clear, clean, openly democratic process.
at the same time, the results were available the same day the election closed. by early evening the outcome was clear. and this is with simple paper ballots counted by hand.
finally, and even more amazingly, next time there are elections for president, i will be able to vote. all you need to do to qualify is live here for five years (i admit i've not checked the details, but that is what i understand). on the other hand, chileans abroad cannot vote.
oh, and voting is compulsory, but only if you register. registering is optional. which seems like a good compromise to me.
(generally i seem to complain about this bloody country, but they deserve a huge amount of credit for the way they run elections. i wish it were this clear and involved as much mass-participation in the uk).
it depends very much on the job, for two reasons. first, you probably aren't going to be writing code all the time. you might spend a lot of time doing design work, requirements gathering, use cases, yadda yadda, or testing, or just sitting in meetings. second, work loads vary by huge amounts, depending on the company, position, etc.
my previous job ended up with 6 day weeks, working over 12 hours a day (i would guess, on average - at times we were there for 24 hours or more). most of that was writing code (with no design, etc etc). no way did i write code in my "spare time" then (and, even when i left, after having had a minor breakdown, for months i had no desire to program).
live and learn - my next (current job) i negotiated as shift working, with longish hours for 8 days, then 6 off. on average, i work average hours, but with (1) less room to work extra hours and (2) lots of time between shifts for my own projects. the employer cares more about design/process and it also pays more than twice as much (but that's due to rather odd circumstances - i live in s. america).
so, in the end, it comes down to what you make of it. you can get sunk into a project (like i was) where you're "responsible for saving everyone's jobs" (the world didn't end when i left, though), and it can ruin your life. or you can take a bit more control and have plenty of time (and energy) left for what you want to do.
(my current project is a little programming language, a bit like forth).
other people have commented that you get to see more than that due to rotation of the earth during the night, and movement of th eearth around the sun during the year.
however, there's a more fundamental reason why this is largely unimportant - the universe doesn't have a special direction, it's pretty much the same everywhere. so while you might not be able to see a certain object from a certain telescope, you can see another one pretty much like it.
there are exceptions, of course. if you're looking at objects in our galaxy then you may need to use a certain telescope, because the position of the galaxy relative to the earth is pretty much fixed. so for "nearby" objects it may be important. also, at the extreme opposite, observations of large scale fluctations in the very early universe (effectively observing *very* distance objects) may require all-sky observations.
but for many interesting objects - other galaxies, quasars, radio galaxies, etc there's no real loss to being restricted to one particular direction.
Google somehow (probably a though a CSS hack) manages to substitute a 1x1 white pixel .GIF file for the page if you try to print it or copy its location.
the page image itself is set as the background image - if you look at the page source (via the "view" menu in the toolbar, search for the embedded CSS fragment "background-image:url") you can get the URL, load it explicitly, and get direct access.
what crap.
he's providing a useful service - collating information and providing access to it. how is this any different to a newspaper or magazine's articles.
in this "digital age" anyone is welcome to publish on the net. this guy is doing well because he's good at what he does. more power to him.
go and do something useful yourself instead of wasting your breath telling others how to behave.
the telescope/instrument control systems are/will not be linux, as far as i know, but the data will be archived and reduced on linux (some reduction might also be on suns and max osx).
:o)
i work at ctio, writing iraf software for noao - iraf is multi-platform, but we develop on linux (currently red hat, about to move to fedora, although i also have it running in debian on my laptop - well, i did until yesterday, when i messed up a kernel recompile/install and lost linux completely (it's an x31, so i need to do a network boot to get it back etc etc)
the israelis once assassinated a man by detonating an explosive in his cellphone -- they waited to hear his voice and ...
ah, that must explain their success in having such a peaceful country.
i was so worried about this post i eventually resorted to cormen et al. the section on counting sorts finally let me sleep soundly.
hmmm. i think your c algorithm is O(n^2) too, but only in the limit of large "count" - then the inner loop is O(n) (in practice, count is not that large, and that inner loop has a very small constant, so that loop is proabably not visible in any timing test with reasonable numbers). but i might not have understood what you're doing, and i'm not 100% sure the inner loop is O(n) even then, because the step size is increasing.
i think it's more likely you're seeing a *huge* difference in constant factors. that's very optimized (i'm impressed) c code - of course that also implies it has a whole pile of restrictions (fixed upper limits on values etc).
i doubt any functional solution is going to give you comparable speed to that code - certainly not at the moment. but you're kind-of comparing apples + oranges (it's a bit sneaky that the c code requires an upper bound from the start).
i agree (that write once variables are a feature, not a bug), but it took me a long time to see that and i was trying to relate to how people who haven't used functional programming before might feel. i don't need to tell you it's good, but i think i do need to explain to people who haven't used a functional language that it is ok to find it odd at first (i certainly did).
i'm curious - what's the c version look like? (the haskell version looks ok to me - it seems to be about n^2 according to timings, and i think that's what you'd expect because the number of tests goes up more-or-less as n).
thanks. another good book (that i should have mentioned - i forgot about it when i was writing that, but just used it a moment ago) is cousineau + mauny's "the functional approach to programming". it's like an ml/ocaml version of the (relatively) famous "structure and interpretation of computer programs".
Here (Santiago, Chile), I get an externally visible address for my cable modem, but it's via DHCP, so can change (in practice, it does so very rarely, so I use it as webmail server; my public site is with a colo in the USA). I can't remember what the contract says, but I wouldn't be surprised if I was only supposed to have one machine connected (that was the case in the UK, when I was there). However, nobody has complained about my using NAT (I have a Linux box as firewall with two internal subnets - one has the Linux box with a Windows machine, the other is connected to a wireless hub; the wireless subnet is filtered by MAC with *no* direct local access to the firewall machine).
Hi, Do you have a reference? (I just started work for NOAO in their data products group, so I'm curious what you're doing).
Going to ADASS?
Cheers,
Andrew
After something similar (but due to bad management rather than bad economy), I resigned and took a months holiday. At the end of the month I returned, negotiated better working conditions (double time for overtime and working from home) and started again.
However the intense period before - when we were working to meet a deadline - was too much (that holiday was to mentally recover - I did little except sleep and occasionally cry/shout/hit walls). I no longer care much about the company and don't enjoy working there (despite the better conditions).
So I'm looking for another job.
Conclusions, then:
- If you're critical to the business you can dictate terms (reasonably, of course)
- Don't leave it too long before doing something, or the harm may already be done.
If I produce a song and want it freely distributed, what must I do?
Go to creative commons.