Thats usual for RS. Phone them and whinge and it'll appear very rapidly. I live about a mile from an RS trade counter which is the best place to go whinge at.
Neckbeard territory here. It's terribly expensive. I got a grant and paid sod all back in the day for which I am very grateful. If we didn't piss 9.3x10^9 quid up the wall on the festival of running, jumping and getting missiles pointed at us, then quite a few more people would have an education in a few more years...
My only regret is that I should have done medicine instead of EE as I hate the software industry as a whole.
Yes but you don't want to have to reinstall it every 5 minutes when you break something. Please don't say the word QEMU either as it's not the same.
I notice you are from the UK. In the mid 90's we had a surplus of BBC micros from schools available. We used them to test hardware interfaces for micrcontrollers so they didn't blow up expensive dev hardware. If it blew up, you'd throw the beeb in a skip and get another one out of the cupboard.
That's what the Pi is for both with respect to software and hardware. It's a "shitbox" you can trash every 5 minutes.
I think it has the best reputation for comp sci research as the grads are shit and can't get a job anywhere else (no offense guys). Agree they are best for mathematics though.
The best guys we've had came from Nottingham, Brunel, Warwick, Reading and Imperial.
At the same time, it involves less thinking now. You had to understand what you're doing rather than get google to do your homework for you.
People now just have no fucking idea what they are doing unless they can Google it. Even I'm guilty of it and I hate it so much thatI drag my laptop out into the garden where the WiFi doesn't work occasionally so I can think for a bit.
There are two reasons to do a university course: to make money -or- because it's interesting. The latter people are so few and far between that they have to cater for the earlier ones to make the courses viable.
At the risk of getting flamed off the planet, most of the postgrads we've had in are shit and couldn't pass the simple test: "you have one hour to open a file in python and print every alternate line to the screen". They were given a windows 7 box without python installed and had to work it out themselves. Quite scarily the worst ones come from the red brick outfits and the best come from the old technical colleges.
I applaud Cambridge for trying to appeal to the people who find it interesting as these are the people we need badly.
I'd prefer "Software Engineering, My Good Friend" which is a whole different boat to Programming Motherfucker.
"Software Engineering, My Good Friend" would be similar but is not patronising, has no t-shirt shop (pocket protectors perhaps), takes well proven engineering principles and apply them to software and deliver quality working software the first time.
The only problem with "Software Engineering, My Good Friend" is that the marketing and business folk don't find it fashionable any more and would rather use "Programming Motherfucker" or "OMG Agile Twat Fuckwad Gnnnrrrr.....Shitbag" to deliver half working crap instead.
There is no runtime code generation done. Everything is done at compile time. We have a custom written annotation processor that allows pre/post/invariant condition checks throughout the code. That is the only code generation.
The application is modular (similar approach to OSGI) so it loads components at runtime on first request. This allows startup times to be minimal. There approximately 85 components which assemble themselves into the application at runtime on demand. This is not possible with C++ unless you resort to things such as CORBA, COM or at a low level dlopen, all of which multiply complexity. Have you ever tried memory management with CORBA? - it's horrid!
I know you can inject the Boehm GC into C++ for example and use your boxing methodology, but that still leaves plenty of windows open and is just trying to turn C++ into Java (why bother?).
Agreed about MFC/Win32/C++ - whilst the NT kernel is a really elegant piece of engineering, the Win32 subsystem on top is a steaming pile of dung. I'd rather have liked to have seen the OS/2 subsystem survive a little longer...
I've used Qt (commercial) extensively (on QNX) and it's not up to the job at hand here. The support is good but you need it too often.
If I was to pick an alternative to Java for this, I would probably pick Go with GCCGO and not C/C++.
It is pretty cool. There is no whitepaper and I no longer work at the organisation unfortunately so publishing one is not likely. I will write it up in a blog post and post it against this at some point in the future (when time allows). None of the information is covered by proprietary rights in my contract so I can say as much as I like:)
Yes it is. It's an awfully unnecessary abstraction.
There are many more ways which are simpler to solve the problems that virtualization supposedly solves.
Virtualization is just an excuse for an IT department to get away with shitty infrastructure planning, poor product selection and bad quality software.
No you got it right. Both drugs and men. That's what 40 years with my great grandmother and the 60's did to him. At least we didn't get war stories (he hid during the wars).
The application is 52Mb. The data varies between 400Mb and 1Tb per node depending on the installation size and imaging data.
It is 100% uptime. It's up whilst it is turned on. It is designed so that a node failure and network failure is not important (eventual consistency) - in fact the nodes regularly power fail when they are taken offline to be moved around (some are attached to medical imaging platforms). It is not one single monolithic system - it's fully distributed.
Erlang gets the wrong end of the stick. We evaluated it and threw it out. The conclusion is that the language/VM doesn't matter. The architecture, data model and communication medium does.
No it's definitely faster if you know what you are doing. The reason C++ is "fast" is that you can easily sacrifice clean interfaces and modularity for raw performance i.e. by using raw memory and pointers etc. The moment you throw that away to build clean interfaces and modularity in (which is essential on larger projects like ours), your performance advantage goes out of the window. We're not doing it wrong - we're leveraging the right technology. It's easier to make serious mistakes in C++ as well and the additional checks required to verify that they are not being made are expensive. In Java, most of this is handled at compile time (g++ checks+valgrind are not sufficient btw).
Regarding downtime. Consider CAP theorem. We use a PAXOS consensus algorithm based protocol between nodes and our own event driven message-oriented container which runs inside the client process. Effectively the system, per-installation is a big message bus. There is no central point of failure. There are no servers to fail. If a single node is up, the system is operational. Scalability comes from CAP theorem - we sacrificed C (consistency) yet apply P (partition tolerance) and A (availability). We have unique reliability requirements which means we don't use a COTS container like Tomcat, Glassfish or Jetty which is what you are most likely used to.
1.9 million lines is due to the complexity of the product - the task it is required to do is not easy to visualize, is processing heavy and is complex. We also have about 2.9 million lines of jUnit and selenium RC tests. It's modular and well maintained as it's built by people who know what they're doing.
I earn plenty thank you.
This is a proper software engineering project, not a startup, internet fad, cost cutting low-rate business.
Well considering the existing causality data is based on picking the "regression analysis" that matches the desired result for marketing healthcare process, I'd say my data was probably more useful.
Please don't forget that stats do not include predispositions either genetically (born with it) or mentally (doomed self to it).
Note: I worked with stats and medical data for years...
I'm currently waiting for someone to build me a new workstation after a disk failure so I have nothing better to do than to sit here and get paid to do this. Life is bliss.
My point is there is no equal. We couldn't have delivered it with Python, Ruby, C, C++ or any other platform out there. Our competitors were struggling along adding features and dealing with maintenance issues with C++/MFC/Win32. It took us less time to build from scratch than it took them to get an interim release out of the door. Survival of the fittest.
You said that the UI sucked. Based on your previous comments, I assumed you were naive enough to consider that to be appearance. I do not claim to be a usability expert (I doubt that there is such a thing as it's subjective and look where 'usability experts have left us': unity, metro, gnome 3). We had three experts from the target market and three separate input paradigms based on whether or not you had a probe or instrument in your hand or whether you just sat behind a normal workstation. None of the UI's "sucked" for the users either with respect to usability or appearance.
Slashdot Mirror
Thats usual for RS. Phone them and whinge and it'll appear very rapidly. I live about a mile from an RS trade counter which is the best place to go whinge at.
Neckbeard territory here. It's terribly expensive. I got a grant and paid sod all back in the day for which I am very grateful. If we didn't piss 9.3x10^9 quid up the wall on the festival of running, jumping and getting missiles pointed at us, then quite a few more people would have an education in a few more years...
My only regret is that I should have done medicine instead of EE as I hate the software industry as a whole.
Yeah - it's worth at least 4 pints second hand!
He's probably from oop-norf. They have a different type of organ to the liver up there.
Yes but you don't want to have to reinstall it every 5 minutes when you break something. Please don't say the word QEMU either as it's not the same.
I notice you are from the UK. In the mid 90's we had a surplus of BBC micros from schools available. We used them to test hardware interfaces for micrcontrollers so they didn't blow up expensive dev hardware. If it blew up, you'd throw the beeb in a skip and get another one out of the cupboard.
That's what the Pi is for both with respect to software and hardware. It's a "shitbox" you can trash every 5 minutes.
I think it has the best reputation for comp sci research as the grads are shit and can't get a job anywhere else (no offense guys). Agree they are best for mathematics though.
The best guys we've had came from Nottingham, Brunel, Warwick, Reading and Imperial.
At the same time, it involves less thinking now. You had to understand what you're doing rather than get google to do your homework for you.
People now just have no fucking idea what they are doing unless they can Google it. Even I'm guilty of it and I hate it so much thatI drag my laptop out into the garden where the WiFi doesn't work occasionally so I can think for a bit.
I think you are talking from the wrong end of your body. It's a general-purpose computer.
It is relatively open, is well documented, has an OS and a toolchain, it's inexpensive and reliable. That's all you need as an educational tool.
What do you think we need as an educational tool instead?
There are two reasons to do a university course: to make money -or- because it's interesting. The latter people are so few and far between that they have to cater for the earlier ones to make the courses viable.
At the risk of getting flamed off the planet, most of the postgrads we've had in are shit and couldn't pass the simple test: "you have one hour to open a file in python and print every alternate line to the screen". They were given a windows 7 box without python installed and had to work it out themselves. Quite scarily the worst ones come from the red brick outfits and the best come from the old technical colleges.
I applaud Cambridge for trying to appeal to the people who find it interesting as these are the people we need badly.
Unfortunately "Linux for Morons" is the only thing likely to grow market share as most humans are morons.
I dont blame them really - for most people, it's just another appliance.
Dull makes stuff that works. Writing code is not all unicorns shitting rainbows.
That will wear out quicker than the display cable on an 80's digital diary...
I'd prefer "Software Engineering, My Good Friend" which is a whole different boat to Programming Motherfucker.
"Software Engineering, My Good Friend" would be similar but is not patronising, has no t-shirt shop (pocket protectors perhaps), takes well proven engineering principles and apply them to software and deliver quality working software the first time.
The only problem with "Software Engineering, My Good Friend" is that the marketing and business folk don't find it fashionable any more and would rather use "Programming Motherfucker" or "OMG Agile Twat Fuckwad Gnnnrrrr.....Shitbag" to deliver half working crap instead.
There is no runtime code generation done. Everything is done at compile time. We have a custom written annotation processor that allows pre/post/invariant condition checks throughout the code. That is the only code generation.
The application is modular (similar approach to OSGI) so it loads components at runtime on first request. This allows startup times to be minimal. There approximately 85 components which assemble themselves into the application at runtime on demand. This is not possible with C++ unless you resort to things such as CORBA, COM or at a low level dlopen, all of which multiply complexity. Have you ever tried memory management with CORBA? - it's horrid!
I know you can inject the Boehm GC into C++ for example and use your boxing methodology, but that still leaves plenty of windows open and is just trying to turn C++ into Java (why bother?).
Agreed about MFC/Win32/C++ - whilst the NT kernel is a really elegant piece of engineering, the Win32 subsystem on top is a steaming pile of dung. I'd rather have liked to have seen the OS/2 subsystem survive a little longer...
I've used Qt (commercial) extensively (on QNX) and it's not up to the job at hand here. The support is good but you need it too often.
If I was to pick an alternative to Java for this, I would probably pick Go with GCCGO and not C/C++.
Fair point!
It is pretty cool. There is no whitepaper and I no longer work at the organisation unfortunately so publishing one is not likely. I will write it up in a blog post and post it against this at some point in the future (when time allows). None of the information is covered by proprietary rights in my contract so I can say as much as I like :)
Yes it is. It's an awfully unnecessary abstraction.
There are many more ways which are simpler to solve the problems that virtualization supposedly solves.
Virtualization is just an excuse for an IT department to get away with shitty infrastructure planning, poor product selection and bad quality software.
No you got it right. Both drugs and men. That's what 40 years with my great grandmother and the 60's did to him. At least we didn't get war stories (he hid during the wars).
The competitor did have the better product. Ours was a naff embedded+C system. That's why we built this one, so please invert your argument.
The application is 52Mb. The data varies between 400Mb and 1Tb per node depending on the installation size and imaging data.
It is 100% uptime. It's up whilst it is turned on. It is designed so that a node failure and network failure is not important (eventual consistency) - in fact the nodes regularly power fail when they are taken offline to be moved around (some are attached to medical imaging platforms). It is not one single monolithic system - it's fully distributed.
Erlang gets the wrong end of the stick. We evaluated it and threw it out. The conclusion is that the language/VM doesn't matter. The architecture, data model and communication medium does.
SWT works with selenium nicely.
Seeing as I made a claim, I'll explain further.
No it's definitely faster if you know what you are doing. The reason C++ is "fast" is that you can easily sacrifice clean interfaces and modularity for raw performance i.e. by using raw memory and pointers etc. The moment you throw that away to build clean interfaces and modularity in (which is essential on larger projects like ours), your performance advantage goes out of the window. We're not doing it wrong - we're leveraging the right technology. It's easier to make serious mistakes in C++ as well and the additional checks required to verify that they are not being made are expensive. In Java, most of this is handled at compile time (g++ checks+valgrind are not sufficient btw).
Regarding downtime. Consider CAP theorem. We use a PAXOS consensus algorithm based protocol between nodes and our own event driven message-oriented container which runs inside the client process. Effectively the system, per-installation is a big message bus. There is no central point of failure. There are no servers to fail. If a single node is up, the system is operational. Scalability comes from CAP theorem - we sacrificed C (consistency) yet apply P (partition tolerance) and A (availability). We have unique reliability requirements which means we don't use a COTS container like Tomcat, Glassfish or Jetty which is what you are most likely used to.
1.9 million lines is due to the complexity of the product - the task it is required to do is not easy to visualize, is processing heavy and is complex. We also have about 2.9 million lines of jUnit and selenium RC tests. It's modular and well maintained as it's built by people who know what they're doing.
I earn plenty thank you.
This is a proper software engineering project, not a startup, internet fad, cost cutting low-rate business.
Well considering the existing causality data is based on picking the "regression analysis" that matches the desired result for marketing healthcare process, I'd say my data was probably more useful.
Please don't forget that stats do not include predispositions either genetically (born with it) or mentally (doomed self to it).
Note: I worked with stats and medical data for years...
Actually no. I think my views on the matter are summed up nicely here: http://whatupdave.tumblr.com/post/1170718843/leaving-net
I think you nailed it here.
I'm currently waiting for someone to build me a new workstation after a disk failure so I have nothing better to do than to sit here and get paid to do this. Life is bliss.
My point is there is no equal. We couldn't have delivered it with Python, Ruby, C, C++ or any other platform out there. Our competitors were struggling along adding features and dealing with maintenance issues with C++/MFC/Win32. It took us less time to build from scratch than it took them to get an interim release out of the door. Survival of the fittest.
You said that the UI sucked. Based on your previous comments, I assumed you were naive enough to consider that to be appearance. I do not claim to be a usability expert (I doubt that there is such a thing as it's subjective and look where 'usability experts have left us': unity, metro, gnome 3). We had three experts from the target market and three separate input paradigms based on whether or not you had a probe or instrument in your hand or whether you just sat behind a normal workstation. None of the UI's "sucked" for the users either with respect to usability or appearance.