"Now while downloading a hundred files to your desktop won't automatically execute them, Microsoft's position is that a secondary attack could execute them for you."
With hundreds of files on your desktop, what are the odds you'd hit one when you are just blanking out a selection, or deleting them, or frustratingly smack your mouse for [whatever reason] Okay...why in @#$@'s name would you make your download vector your desktop?
I have a nice little directory (folder for the uninitiated) that holds all of my downloads.
Charleton Heston was a HAM - I never saw any fragility or imperfections of character in his performances; even when he was the underdog, he didn't come across as the underdog.
Now, Will's performance in IAmLegend - wow. The despair was palpable...
While entertaining in that 'omg-this-is-making-me-cringe' kind of way, Tron, Hackers, and other such dross have been shown conclusively to lower the viewer's IQ by at least 10 points for several hours after viewing.
For a person who has viewed these movies at least once, it has also been shown to lower the viewer's IQ 20 points 1 hour prior to viewing, in anticipation - as the mind recoils in shock and dread.
That being said, if you have the stamina of the Mystery Science Theatre 3000 guy, it can be entertaining to make witty repartee during the time of exposure, and also serves to cushion the negative effects on your IQ.
To expand on your comment, I would say there are certain tasks that are suitable for all the languages listed (and then some):
Assembly - embedded systems/drivers for a given platform.
C - larger systems programming tasks that need to be portable -- too big to be effectively built in Assembly without going crazy, too small for C++; you can embed Assembly instructions in C as needed - so you can have your cake and eat it too.
C++ - for complex systems requiring near-real-time response (e.g. multiplayer video game).
Ada - for complex systems requiring real-time control (e.g. avionics) without error.
Java - for complex systems not requiring real-time control or response (e.g. GUI Office Suite that interfaces to user via http).
Python - for simple or complex systems not requiring real-time control - but where speed of deployment/refactoring is more important than processing speed (e.g. embedded control language, or specialized control/reporting interface tied to a changing back-end system).
It is thus irrelevant to talk about 'java is faster than C++' or any similar argument, without including a complete evaluation of the factors involved in a given project. The factors would include, the time required to complete the project, as well as performance factors.
We can certainly agree that building a GUI http interface using Assembly would not be optimal, given the ease of using Java to do the same thing. Of course, there may be an instance when it would be more desirable (perhaps building it in an embedded system with severe limitations on memory or performance).
As a developer, you should be able to reach into your bag and pull out whatever tool is best suited for a given job. One-trick ponies end up in the glue factory.
Religious wars are counter productive. We should frame the discussion to examine the practicalities of a given set of projects.
In Computer Science terms, atomic operations are ones which prevent the 'he who writes last - wins' syndrome.
In multiprocessor architectures the use of resource locking normally does the job, although one of the drawbacks is forcing other threads of execution to wait until the lock is released. There are non-locking and wait-free methods that use low level atomic operations (like 'compare and save') and generally use data structures that support these methods (stacks, queues, sets and hashes).
Rarely, if ever, is the whole continent covered with clouds. If we are talking about a system to pump energy into the grid for far flung cities, why not build a dispersed grid of these solar collectors?
This way, the ones that are socked-in could be backed up by the ones that are not.
This could also be said for wind and hydro power, which to a certain extent are all dependent upon mother nature.
Not to put words in his mouth, but I think the grandparent post was making the point that it is the 'fit and finish' (e.g. integration) of the hardware and the software that is the differentiating factor:
Apple is selling exactly the same hardware for twice as much. The only difference is software.
While he may not be completely accurate, I think he makes a good point when you read between the lines.
Unless an astrophycisst (sic - lazy) has actual numbers as to what % of the total matter is "dark"... There are no such thing as actual numbers at astronomical distances.
The distance to a given star can only be accurately triangulated for relatively short distances (several light years). Beyond that astronomers use 'standard candles' - such as cepheid variable stars and supernovae - to estimate the distance to a star using the known luminosity of these adjacent objects.
That being said, if we determine that our observations of the luminosity of the given objects is incorrect due to this absorption then that throws all of the calculations out the window. Not only do they have more mass than anticipated, they are also closer to us than previously thought.
I think much of the tension between adherents of Ada and other compiled languages, and the newer languages has to do with strictures and philosophy surrounding data types and the overhead, or lack thereof, to manage same.
I do agree there is a place and applications for strongly typed languages. But I also believe there is a happy medium that can be obtained for most other programs. Python, for example, uses 'Duck' typing - which is considered type-safe, in comparison to other languages (such as C and C++). Simply put, 'Duck' typing makes no assumptions (in fact has no knowledge) of the type of a given variable - instead it assumes that there is a method available to perform a given action on the variable, and if not - raises a runtime error - which can be handled by the application. Python's implementation of 'Duck' typing is also strongly typed, as a result. This has helped me immensely - things that I used to be able to do (even if by accident or omission) in Perl or C, are impossible in Python.
This allows a developer to safely code very quickly, and refactor and maintain code easily. In the future, I could see where more languages handle much of the duplication required of traditional compiled languages behind the scenes in ways that are fool proof; we've already seen this with garbage collection systems used in various languages to avoid memory leak conditions - and as languages get more mature there will be opportunities in other areas.
As a result I think Ada and the like will remain niche players, and more advanced languages will evolve to provide secure and viable code approaching the level of consistency and error free operation of the old languages.
My original comment was an oblique response to this agenda put forward in the article:
Computer geeks who can afford learning a language simply because it is cool have now access to an Ada compiler. My hope is that some day they will grow into positions in IT industry and bring Ada into the widespread use that the language deserves, without having to convince anybody. -
Joachim Schüeth
I disagree with the author: there is no reason to bring Ada and similar old languages into widespread use. The energy needs to be put into finding ways to make programmers more efficient by leveraging the computer to move forward - much as we have seen with python and similar languages. Just as assemblers gave way to compilers, and structured languages gave way to OO and hybrid languages, error creation as a symptom of the foibles of a given language will be taken out of the programmers hands.
If I could type 90 wpm and never make a mistake, I'd be using Ada or Fortran today.
I can't, so I don't.
For me it is how fast I can produce what the customer/user desires that matters - and more importantly, how fast I can change it - so I use Python - with the option (as yet unneeded) to build C/C++ modules for that language for slow bits. If a bug pops out of my code, I can easily squash it; more difficult with a compiled language.
Back in the reel-to-reel tape days a similar approach (can't recall if it used iron oxide, or laser printer paper to capture the image) was used to visually read the contents of damaged or worn tapes that could not be read from the tape reader.
I was thinking the same thing. But then I thought about all the in-house pod-casts and videos etc that have been coming out from our executives in the last few years. Would be kind of hard to get the message, unless you could lip read.
On the other hand, most of this stuff is transcribed somewhere else anyway - so you can read what they said - without having to wade through a bunch of blather.
Additionally, shouldn't his team have a spare machine for just such emergencies? This is a prime reason I think a certain percentage of machines should be purchased, as opposed to being rented. The machines that are purchased can go into a pool of spares when the user in question upgrades, that can continue to serve in such emergencies.
Good points, but Cisco doesn't make servers. I hate to break it to you but Cisco's AXP server module turns Cisco switches into application servers running a hardened version of Linux.
security through obscurity - don't provide patches at all, and never release the source code. This will make it much harder for script kiddies.
It should be noted that Microsoft Research is aligned with the CMU Computer Science department - vis a vis the Microsoft Carnegie Mellon Center for Computational Thinking. That is either ironic, or obvious depending on your viewpoint (either it can fuel Microsoft to do a more thorough job of releasing early, often and being more transparent, or more likely, it will confirm Microsoft's assertion that security through obscurity is the way to - continue - going).
My first choice would have been MIT, but I didn't have the where-with-all (money, ability to relocate etc) that would have required (I was married with children when I decided to go back to school).
Instead I went to the local state university, and to further save money went to a local community college that had a 2+2 program agreement with the university (freshman and sophomore years at the community college with full transfer credits, and junior and senior years at the university).
The nice thing about this approach was I avoided the 300+ student lecture hall courses (I started at the university initially - and that was a big pain...waiting in the halls with 150 other people to see the professor...not the most efficient use of my time - and another major reason I transfered and did my first two years at the community college). When I returned to the university, I was in smaller upper-division classes...small classes equate to a better learning environment in most cases - and a good chunk of the lower division classes at the university were used to weed-out uncommitted folks (the basic programming course and the unix shell programming courses served this purpose in the CS department - classes that started with 70-50 people at the beginning of the semester ended with 30-20).
Additionally, the university's computer science department chair was an MIT professor, and her husband as well; and both of them taught some excellent and memorable courses (computer architecture, unix system programming, and relational databases) - I still have the books from these, and reference these concepts on an almost daily basis in my job.
The course work also included assembly language programming, and a logic design course (hardware circuit design - using a simulation program to validate your work) - important to understand the underlying functionality of that 4GL framework you are using. Of course it also included English, history, technical writing, government, economics, higher level math (calc I, II, III, linear algebra, statistics), physics and other courses designed to give you a well rounded undergraduate education.
I would recommend this approach if you weren't born with a silver spoon in your mouth, can't get scholarships, grants, or take on large amounts of debt (I am still paying off my college loans --- I did change my major from aviation to computer science, so you might not incur such high costs if you know what you want and go for it straight away).
I also recommend doing some research -- find out what courses are available and required for the CS program -- as a minimum I would expect an equal measure of math and computer courses (and there may be software/hardware branches - so think about what you want to pursue) in addition to your standard requirements If the CS courses are all fluff (focusing mainly on data processing - e.g. 4GL programming frameworks/java and SQL/DB2) with little deep theoretical courses (such as compiler design, boolean algebra, systems architecture and systems programming etc) - then you should steer clear; you want the broadest possible exposure to various computer science concepts in the undergraduate course imho.
Additionally, look at the instructors - are they professors of computer science, or just teaching it? Do they come from universities that are known for computer science (MIT, Columbia etc) specifically what is their lineage - did they work with giants in the discipline - or have little understanding of the culture? Remember - you are paying for this - be sure to get your money's worth.
I got lucky - some of my key instructors were themselves instructed by some really excellent professors - and their enthusiasm and depth of knowledge showed - and helped me understand the subjects at a deeper level than I would have achieved otherwise.
Subject of the Emmy and Peabody Award-winning 2004 HBO film "Something the Lord Made".
Not only was Vivien Thomas challenging the professor/student barrier, he was also challenging the racial barrier many years before the emergence of the popular civil rights movement. He was a pioneer on many levels.
At least his contributions were recognized eventually with better pay in the 1940s, and an honoray doctorate and permanent faculty status bestowed in 1976 (three years before his retirement).
There are many more who's stories we'll never know.
My company took another approach. It is a subsidiary of a larger holding entity. That entity also has an 'IT' subsidiary that is supposed to provide all IT services - but that group is really geared towards enterprise data-center level applications, and can not effectively handle either one-sy-two-sy type work, or conversely larger projects that required a great deal of change/integration with other systems and 24/7 support. Project after project failed, or was delivered in a crippled state - and not in time to take advantage of limited windows of opportunity in many cases.
We set up a small team of developers dedicated to doing reporting and internal tools (I am one of the developers - and we are not super users - I am a CS/*nix guy) for the operations group as part of a reorganization. This grew organically from the technical engineering and system administration groups. Where IT can't or won't build a given application, we step in. We also interface with the end/super users - and fold in their requirements into the projects we are pursuing.
This does two things: 1. It keeps development and testing on dedicated development systems, and deployment on our dedicated production systems...no worries of a rogue superuser cratering the network.
2. It satisfies the needs of the business to accomplish those tasks that fall through the cracks in the IT process and scope. Management likes not getting the run-around when they need something fast.
This probably wouldn't work for all organizations - since the talent and resources to do development in a safe manner (dedicated development/test/production systems, version control, knowledge of a wide range of application and network issues surrounding client/server applications etc) is probably not available. One solution in such an environment might be to dedicate some IT resources for this purpose exclusively.
Get in the trenches, live with the users, eat some of that dog food you're serving.
There is a certain tyranny of expertise - particularly in academia. No matter how well researched, thought out, or tested a particular product is (whether it be object or manuscript), it will be snubbed unless the author/inventor has 'Doctor' after his/her name.
I used to think the institutions of higher learning were composed of open minded people - until I went to school. With rare exception this is not the case - dogma wins out over discourse. The unwitting student stumbles into this minefield of vested interests - the teacher actively attempts to suppress the heretical concepts, or more commonly brushes them under the rug with little comment and much condescension.
While professors challenge their students to think critically and with an open mind, they should also take that same advise to heart.
I'm thinking 'Dumbcubator' - particularly if the tech students interact in any significant way with the business students.
The business students won't gain anything from the lash-up because they won't be able to understand what the techies are saying, and the tech students IQs will drop by double digits from listening to inane golfing stories.;)
Slashdot Mirror
With hundreds of files on your desktop, what are the odds you'd hit one when you are just blanking out a selection, or deleting them, or frustratingly smack your mouse for [whatever reason] Okay...why in @#$@'s name would you make your download vector your desktop?
I have a nice little directory (folder for the uninitiated) that holds all of my downloads.
That is when the tools would come out...and the door would come off the hinges.
My question would be, do you have to pay to go in, or get out?
Charleton Heston was a HAM - I never saw any fragility or imperfections of character in his performances; even when he was the underdog, he didn't come across as the underdog.
Now, Will's performance in IAmLegend - wow. The despair was palpable...
While entertaining in that 'omg-this-is-making-me-cringe' kind of way, Tron, Hackers, and other such dross have been shown conclusively to lower the viewer's IQ by at least 10 points for several hours after viewing.
For a person who has viewed these movies at least once, it has also been shown to lower the viewer's IQ 20 points 1 hour prior to viewing, in anticipation - as the mind recoils in shock and dread.
That being said, if you have the stamina of the Mystery Science Theatre 3000 guy, it can be entertaining to make witty repartee during the time of exposure, and also serves to cushion the negative effects on your IQ.
To expand on your comment, I would say there are certain tasks that are suitable for all the languages listed (and then some):
Assembly - embedded systems/drivers for a given platform.
C - larger systems programming tasks that need to be portable -- too big to be effectively built in Assembly without going crazy, too small for C++; you can embed Assembly instructions in C as needed - so you can have your cake and eat it too.
C++ - for complex systems requiring near-real-time response (e.g. multiplayer video game).
Ada - for complex systems requiring real-time control (e.g. avionics) without error.
Java - for complex systems not requiring real-time control or response (e.g. GUI Office Suite that interfaces to user via http).
Python - for simple or complex systems not requiring real-time control - but where speed of deployment/refactoring is more important than processing speed (e.g. embedded control language, or specialized control/reporting interface tied to a changing back-end system).
It is thus irrelevant to talk about 'java is faster than C++' or any similar argument, without including a complete evaluation of the factors involved in a given project. The factors would include, the time required to complete the project, as well as performance factors.
We can certainly agree that building a GUI http interface using Assembly would not be optimal, given the ease of using Java to do the same thing. Of course, there may be an instance when it would be more desirable (perhaps building it in an embedded system with severe limitations on memory or performance).
As a developer, you should be able to reach into your bag and pull out whatever tool is best suited for a given job. One-trick ponies end up in the glue factory.
Religious wars are counter productive. We should frame the discussion to examine the practicalities of a given set of projects.
In Computer Science terms, atomic operations are ones which prevent the 'he who writes last - wins' syndrome.
In multiprocessor architectures the use of resource locking normally does the job, although one of the drawbacks is forcing other threads of execution to wait until the lock is released. There are non-locking and wait-free methods that use low level atomic operations (like 'compare and save') and generally use data structures that support these methods (stacks, queues, sets and hashes).
Rarely, if ever, is the whole continent covered with clouds. If we are talking about a system to pump energy into the grid for far flung cities, why not build a dispersed grid of these solar collectors?
This way, the ones that are socked-in could be backed up by the ones that are not.
This could also be said for wind and hydro power, which to a certain extent are all dependent upon mother nature.
While he may not be completely accurate, I think he makes a good point when you read between the lines.
The distance to a given star can only be accurately triangulated for relatively short distances (several light years). Beyond that astronomers use 'standard candles' - such as cepheid variable stars and supernovae - to estimate the distance to a star using the known luminosity of these adjacent objects.
That being said, if we determine that our observations of the luminosity of the given objects is incorrect due to this absorption then that throws all of the calculations out the window. Not only do they have more mass than anticipated, they are also closer to us than previously thought.
I do agree there is a place and applications for strongly typed languages. But I also believe there is a happy medium that can be obtained for most other programs. Python, for example, uses 'Duck' typing - which is considered type-safe, in comparison to other languages (such as C and C++). Simply put, 'Duck' typing makes no assumptions (in fact has no knowledge) of the type of a given variable - instead it assumes that there is a method available to perform a given action on the variable, and if not - raises a runtime error - which can be handled by the application. Python's implementation of 'Duck' typing is also strongly typed, as a result. This has helped me immensely - things that I used to be able to do (even if by accident or omission) in Perl or C, are impossible in Python.
This allows a developer to safely code very quickly, and refactor and maintain code easily. In the future, I could see where more languages handle much of the duplication required of traditional compiled languages behind the scenes in ways that are fool proof; we've already seen this with garbage collection systems used in various languages to avoid memory leak conditions - and as languages get more mature there will be opportunities in other areas.
As a result I think Ada and the like will remain niche players, and more advanced languages will evolve to provide secure and viable code approaching the level of consistency and error free operation of the old languages.
My original comment was an oblique response to this agenda put forward in the article:
I disagree with the author: there is no reason to bring Ada and similar old languages into widespread use. The energy needs to be put into finding ways to make programmers more efficient by leveraging the computer to move forward - much as we have seen with python and similar languages. Just as assemblers gave way to compilers, and structured languages gave way to OO and hybrid languages, error creation as a symptom of the foibles of a given language will be taken out of the programmers hands.
If I could type 90 wpm and never make a mistake, I'd be using Ada or Fortran today.
I can't, so I don't.
For me it is how fast I can produce what the customer/user desires that matters - and more importantly, how fast I can change it - so I use Python - with the option (as yet unneeded) to build C/C++ modules for that language for slow bits. If a bug pops out of my code, I can easily squash it; more difficult with a compiled language.
YMMV
Glad you found it mildly amusing. Anything to please the crowd.
I can't imagine using the words concise and Ada in the same sentence.
Constricted - maybe. Painful - most certainly.
I would also add a satellite data link for those times when you are out in the wilderness with no wifi signals...
Back in the reel-to-reel tape days a similar approach (can't recall if it used iron oxide, or laser printer paper to capture the image) was used to visually read the contents of damaged or worn tapes that could not be read from the tape reader.
I was thinking the same thing. But then I thought about all the in-house pod-casts and videos etc that have been coming out from our executives in the last few years. Would be kind of hard to get the message, unless you could lip read.
On the other hand, most of this stuff is transcribed somewhere else anyway - so you can read what they said - without having to wade through a bunch of blather.
Additionally, shouldn't his team have a spare machine for just such emergencies? This is a prime reason I think a certain percentage of machines should be purchased, as opposed to being rented. The machines that are purchased can go into a pool of spares when the user in question upgrades, that can continue to serve in such emergencies.
There is a fourth choice -
security through obscurity - don't provide patches at all, and never release the source code. This will make it much harder for script kiddies.
It should be noted that Microsoft Research is aligned with the CMU Computer Science department - vis a vis the Microsoft Carnegie Mellon Center for Computational Thinking. That is either ironic, or obvious depending on your viewpoint (either it can fuel Microsoft to do a more thorough job of releasing early, often and being more transparent, or more likely, it will confirm Microsoft's assertion that security through obscurity is the way to - continue - going).
My first choice would have been MIT, but I didn't have the where-with-all (money, ability to relocate etc) that would have required (I was married with children when I decided to go back to school).
Instead I went to the local state university, and to further save money went to a local community college that had a 2+2 program agreement with the university (freshman and sophomore years at the community college with full transfer credits, and junior and senior years at the university).
The nice thing about this approach was I avoided the 300+ student lecture hall courses (I started at the university initially - and that was a big pain...waiting in the halls with 150 other people to see the professor...not the most efficient use of my time - and another major reason I transfered and did my first two years at the community college). When I returned to the university, I was in smaller upper-division classes...small classes equate to a better learning environment in most cases - and a good chunk of the lower division classes at the university were used to weed-out uncommitted folks (the basic programming course and the unix shell programming courses served this purpose in the CS department - classes that started with 70-50 people at the beginning of the semester ended with 30-20).
Additionally, the university's computer science department chair was an MIT professor, and her husband as well; and both of them taught some excellent and memorable courses (computer architecture, unix system programming, and relational databases) - I still have the books from these, and reference these concepts on an almost daily basis in my job.
The course work also included assembly language programming, and a logic design course (hardware circuit design - using a simulation program to validate your work) - important to understand the underlying functionality of that 4GL framework you are using. Of course it also included English, history, technical writing, government, economics, higher level math (calc I, II, III, linear algebra, statistics), physics and other courses designed to give you a well rounded undergraduate education.
I would recommend this approach if you weren't born with a silver spoon in your mouth, can't get scholarships, grants, or take on large amounts of debt (I am still paying off my college loans --- I did change my major from aviation to computer science, so you might not incur such high costs if you know what you want and go for it straight away).
I also recommend doing some research -- find out what courses are available and required for the CS program -- as a minimum I would expect an equal measure of math and computer courses (and there may be software/hardware branches - so think about what you want to pursue) in addition to your standard requirements If the CS courses are all fluff (focusing mainly on data processing - e.g. 4GL programming frameworks/java and SQL/DB2) with little deep theoretical courses (such as compiler design, boolean algebra, systems architecture and systems programming etc) - then you should steer clear; you want the broadest possible exposure to various computer science concepts in the undergraduate course imho.
Additionally, look at the instructors - are they professors of computer science, or just teaching it? Do they come from universities that are known for computer science (MIT, Columbia etc) specifically what is their lineage - did they work with giants in the discipline - or have little understanding of the culture? Remember - you are paying for this - be sure to get your money's worth.
I got lucky - some of my key instructors were themselves instructed by some really excellent professors - and their enthusiasm and depth of knowledge showed - and helped me understand the subjects at a deeper level than I would have achieved otherwise.
ymmv
Subject of the Emmy and Peabody Award-winning 2004 HBO film "Something the Lord Made".
Not only was Vivien Thomas challenging the professor/student barrier, he was also challenging the racial barrier many years before the emergence of the popular civil rights movement. He was a pioneer on many levels.
At least his contributions were recognized eventually with better pay in the 1940s, and an honoray doctorate and permanent faculty status bestowed in 1976 (three years before his retirement).
There are many more who's stories we'll never know.
My company took another approach. It is a subsidiary of a larger holding entity. That entity also has an 'IT' subsidiary that is supposed to provide all IT services - but that group is really geared towards enterprise data-center level applications, and can not effectively handle either one-sy-two-sy type work, or conversely larger projects that required a great deal of change/integration with other systems and 24/7 support. Project after project failed, or was delivered in a crippled state - and not in time to take advantage of limited windows of opportunity in many cases.
We set up a small team of developers dedicated to doing reporting and internal tools (I am one of the developers - and we are not super users - I am a CS/*nix guy) for the operations group as part of a reorganization. This grew organically from the technical engineering and system administration groups. Where IT can't or won't build a given application, we step in. We also interface with the end/super users - and fold in their requirements into the projects we are pursuing.
This does two things:
1. It keeps development and testing on dedicated development systems, and deployment on our dedicated production systems...no worries of a rogue superuser cratering the network.
2. It satisfies the needs of the business to accomplish those tasks that fall through the cracks in the IT process and scope. Management likes not getting the run-around when they need something fast.
This probably wouldn't work for all organizations - since the talent and resources to do development in a safe manner (dedicated development/test/production systems, version control, knowledge of a wide range of application and network issues surrounding client/server applications etc) is probably not available. One solution in such an environment might be to dedicate some IT resources for this purpose exclusively.
Get in the trenches, live with the users, eat some of that dog food you're serving.
There is a certain tyranny of expertise - particularly in academia. No matter how well researched, thought out, or tested a particular product is (whether it be object or manuscript), it will be snubbed unless the author/inventor has 'Doctor' after his/her name.
I used to think the institutions of higher learning were composed of open minded people - until I went to school. With rare exception this is not the case - dogma wins out over discourse. The unwitting student stumbles into this minefield of vested interests - the teacher actively attempts to suppress the heretical concepts, or more commonly brushes them under the rug with little comment and much condescension.
While professors challenge their students to think critically and with an open mind, they should also take that same advise to heart.
solution: Wax fruit/vegetables
You don't have to turn on the fridge, therefore no condensation issues.
I'm thinking 'Dumbcubator' - particularly if the tech students interact in any significant way with the business students.
;)
The business students won't gain anything from the lash-up because they won't be able to understand what the techies are saying, and the tech students IQs will drop by double digits from listening to inane golfing stories.