Hence the paragraph after this in my post. The average user is not going to install bleeding-edge use-at-your-own-risk here-be-dragons beta software just so they can find bugs, and if they do then how many of them will report bugs.
The same is true for Windows. Windows 7 was tested largely by the enthusiasts (and some reporters/reviewers as well). Issues *were* found in this process, and a lot of smooth edges were ironed out.
But what was the hardware coverage for this? Software? Deviations from the base configuration?
The distros and upstream developers need as much coverage as possible to identify bugs and fix the issues. They also need people who are willing to be patient and file bugs, and help the distros and upstream track down the bugs.
Users may also get burned when filing bugs if they don't see any action on the bugs they are interested in, or have filed. But if they don't file them, the developers won't know about them and can't possibly fix them (unless the developers and/or other testers hit a similar issue).
I ended up installing alsa. It would have been nice if it detected that I wouldn't have the sound driver for alsa installed when it overwrote the modprobe configuration, though.
I should really create an oss4 metapackage to help with configuration. However, I have since done a clean install and pulseaudio and the alsa sound drivers are working reasonably well (a lot better than in previous releases).
There are still some issues (e.g. sound in flash disappears after a while, using Adobe's plugin), but as I understand Ubuntu's configuration of pulse is not correct yet; here's hoping that they finally get it right in Lucid.
There are essentially two classes of distribution:
stable (e.g. Debian stable) -- this distribution will not upgrade components until they are settled and working; tends to lag behind on new, shiny features.
latest-and-greatest (e.g. Debian unstable, Ubuntu) -- adds new things and the latest versions of components/drivers without knowing how good they are.
Ubuntu is trying to compete with Windows and Mac, and needs to keep up with evolving features to get the latest functionality (e.g. hot-pluggable input devices in Xorg). It is also focused on meeting requirements of users, so has the option for binary drivers (nvidia, some wifi) so that things can (or should) work.
Exactly. I have been using Karmic from a clean install without major issues, but upgrade killed my sound. Note that I did have OSS4 configured on 9.04 before the upgrade.
One thing I have noticed (and I haven't used other distros to see if this is a common phenomena) is that upgrading Ubuntu is temperamental when it comes to non-standard configurations/customisations (e.g. removing pulseaudio or totem).
Windows: Don't adopt until Service Pack 1 Ubuntu: Don't adopt until 1-2 months after release
Microsoft, Canonical, et. al. are in an interesting position. You need a large number of testers running on a wide variety of hardware (intel, nvidia, ati,...) with a wide variety of configurations (gnome, KDE, modified sound configurations, heavily modified/customised), needs and requirements (a DJ/sound studio will have different requirements to someone who just wants email). If people don't use the OS because it is 'buggy', the bugs are not found and no fixes are released. Hopefully, they will get good coverage over the alpha and beta phases, then a wider adoption with the enthusiasts and early adopters, followed later by your "average" "novice" user.
Plus, Karmic and other Linux distros are not equivalent to Windows; they are equivalent to Windows, Office, Photoshop and a whole host of other applications, all updated and packaged every 6 months (for karmic, main is about 6.7 GiB and universe is in the 25 GiB mark, not sure about multiverse). That's a lot of stuff to ensure is working on everyones machines.
Which is no different to a half-way decent optimiser.
Writing to a pointer does not discard a lot of variables cached in registers; it should discard one if there are no registers free. If it does, the compiler has a very poor registry allocator. It should use one register for a standard loop over contiguous memory for the pointer and continually update that one.
A JIT compiler does have the advantage of being able to do on-the-fly optimisations like you suggest, but the more complex the optimisations, the longer the time needed to generate the jitted code.
The reason iterating through a list is likely to be slower than walking through contiguous memory is that you are likely to require more memory reads with list traversal. Contiguous memory access will keep the data in level 2 cache longer, so the CPU will not be waiting to get the memory.
There are techniques for lists like using a pool allocator -- that is, allocating list items in blocks of n, so you are more likely to keep them in the L2 cache. That is: it all depends on how the data structures are implemented.
Comparing two different data structures is sure to give different performance results. It's like saying a C++ std::vector is faster than a C++ std::list. Also, you need to specify the operations you are performing on the data structures, as they have different performance characteristics (e.g. list has O(1) add and remove, but O(n) random access; vector has O(1) access and but does not guarantee O(1) add/remove (as it may need to allocate memory to expand)).
And also most half-way decent list implementations will block-allocate list nodes in contiguous memory, reusing nodes as they are released. This should yield better allocation performance.
Be aware of the performance characteristics of data structures and algorithms, choosing the best one for the task being performed; perform profiling to see where the slowdowns are and use that to back up data structure/algorithm choices.
Yes - what they are doing is more Chrome-like. They are aiming to reduce visual and UI clutter/redundancy and add better support for things like Vista Glass.
Since it is the "Office UI", they are referring to other office suites (i.e. OpenOffice, KOffice and friends) as "clone products". See the "Can any applications use the license?" section of the link you referenced.
Not sure whether creating wrappable toolbars grouped in tabs that can resize and show/hide text labels depending on size is worthy of protecting, or even that innovative. The application of the UI may have been new, and the research done well, but the UI itself is really an evolution of toolbars, tabs and tabbed dialogs.
Yes. And TFA is taken initially from https://wiki.mozilla.org/Firefox/Sprints/Windows_Theme_Revamp/Direction_and_Feedback, which is discussing the direction of *applications* written for Vista and Windows 7 that don't use the menubar, but use a contextual strip (Windows Explorer) or Office Ribbon (Paint and Wordpad). That paragraph is about the rationale for not showing the menubar on Vista and later in Firefox, not on adding a ribbon to Firefox (it is under a Hiding of the Menubar section).
It seems as though a blogger misread this paragraph, and everyone on the interweb has been taking this as fact, without actually RTFOA (Reading The Friendly Original Article).
From the pcpro article referenced in the/. summary:
"Starting with Vista, and continuing with Windows 7, the menu bar is going away," notes Mozilla in its plans for revamping the Firefox user interface. "[It will] be replaced with things like the Windows Explorer contextual strip, or the Office Ribbon, [which is] now in Paint and WordPad, too."
From the Mozilla page:
"Starting with Vista, and continuing with Windows 7, the menubar is going away. To be replaced with things like the Windows Explorer contextual strip, or the Office Ribbon(now in Paint and Wordpad too). Many apps still retain the menubar as an option to be pinned or to be shown briefly by holding the Alt key."
Note that here they are talking about Vista and Windows 7, not Firefox (and also note the "Many apps..." bit in the last sentence).
Translation: crippled to work differently to BIND8 (causing issues when communicating with a real BIND8 server), or providing additional features that only work with Microsoft DNS.
The point is the implication. They say that freeware implementations of DNS are not malware, but are just as bad. They are using freeware to refer to the open source implementations (are there any freeware implementations of DNS - i.e. ones that you don't have to pay for and don't come with the source code) -- see also how they say that theirs is better because you cannot see the source code.
How much of the problems are due to Ubuntu using bleeding-edge packages? How much of the problems are due to interfacing with binary-only drivers (e.g. nVidia drivers, which are needed for acceleration and 3d graphics; yes, there is nouveau, but that is still a work-in-progress)?
And Greg KH has put in a lot of effort getting the Hyper-V drivers into shape.
Now, lets say that there is some code that the maintainer does not understand. Because they don't have access to the hardware, software (in the case of virtualisation platforms) or technical specs, what should they do?
And yes, it is the *community* (of which Microsoft has become a part). If Microsoft has a vested interest in these drivers, they need to help the community (it is, after all a two-way process).
Microsoft: Here is a binary Linux kernel module for Hyper-V. Someone: But wait... it is using GPL-only kernel interfaces. Microsoft: Oh, er... Grek KH: You need to release the code to be compliant. Microsoft: Here you go. Here is the code for the driver. Microsoft PR: Microsoft has generously donated a lot of code to the Linux kernel under the GPL license. We did this because it is the right thing to do. [Slashdot story #1] [Greg KH spends a lot of time thanklessly getting the code to the point where it is ready for inclusion with the kernel.] [Greg KH sends e-mail to Microsoft asking for help.] [Greg KH posts a summary for the upcoming.32 drivers.] [Slashdot story #2] Microsoft: Ok, we'll help maintain it. Microsoft PR: Microsoft is actively involved with the Linux community.
Yes. That was one thing that I immediately noticed -- they have a scrollbar on a menu (that is not even taking up the whole screen)?!
Some other things I noticed:
1/ In the readability section, the apple one is nice and clear (there is a single product, single focus to the page); the microsoft one -- wtf is the page telling me? That I can get a decent laptop for under $1700?! It also has prominent placement for things like IE8, "Windows Live Mail" and "Windows Live" and other Microsoft products.
Another example: go to the Windows 7 site and you get information about what is new about Windows 7. I personally think that this is designed well (it is on point like the apple one is).
Now go to the Vista site and the main content contains an advertisement for buying a new PC now to upgrade to Windows 7 later, with "Thinking about Windows 7?" below. In fact the main bulk of this page is about Windows 7, or buying a laptop that you can upgrade to Windows 7. WTF does this have to do with Vista?
NOTE: Both of these pages prompt you to "Upgrade your internet experience" if not using IE8 and Silverlight. They also have too many link toolbars and search at the top that eats into the vertical space (especially if viewing it on a widescreen monitor/laptop).
2/ The content below the news section on the microsoft page changes when you mouse over it. This causes the content to be refreshed (resulting in the page height changing in some cases, which can cause the page to jump around, and meaning that it is not easy to link to that information or copy sections of it).
That really annoys me too. Plus, searching for Win32 APIs (on the MSDN search -- which uses whatever the Microsoft search engine is called today -- Live, Bing,...) either gives you results for.NET, or takes you to the WinCE/Mobile API documentation!
Google is the best way as it does not have the broken links (from my observations) and gets you to the right platform.
Plus, I've learned not to rely on what MSDN says -- they even get the API wrong (e.g. it says that the ICreateErrorInfo Set methods take LPCOLESTR when they actually take LPOLESTR arguments).
The applications are there; no need to compile code or anything crazy to install them. Same with other distributions.
What *is* needed is a better way to make this visible to people. Also, add user reviews / ratings / rankings (with some moderation or other system to remove spam and unrelated comments). Add a way to see most popular (today / this week / this month / all time; by download / usage / rating) and "recommended" applications.
Group games by sub-category, so you can see card games, first-person shooters, real-time strategy, etc.
Include screenshots, so people can see what they are like.
Provide links to bug reports (for known/reported) issues as well as forums for help for both Ubuntu (or your favorite distribution of choive) and the application.
Provide "if you like this, people recommend..." and "people using this are also using..."
Provide a way for externally supported applications / versions (e.g. Firefox 3.5/Minefield and Songbird). This could help seeing what applications are popular for future distribution releases.
What would be good is if this were done across all distributions. That way you get a unified experience and can help you in distribution choices (which distribution has better KDE4 support?).
... because it updates *everything* (the operating system and all installed applications that come from the distribution).
And the "cannot tell what updates are required and which are optional" comment in Linux is ridiculous. In the update manager on Ubuntu (checked on 9.04), it clearly shows updates with "Important security updates", "Recommended updates" and "Other updates" listed, with a description of the changes.
I use both flac and ogg. I also only use devices that support these formats - currently the Cowon S9 (a very nice device). Cowon are very good at supporting these formats and have for several years.
Slashdot Mirror
F.U.B.A.R.
Hence the paragraph after this in my post. The average user is not going to install bleeding-edge use-at-your-own-risk here-be-dragons beta software just so they can find bugs, and if they do then how many of them will report bugs.
The same is true for Windows. Windows 7 was tested largely by the enthusiasts (and some reporters/reviewers as well). Issues *were* found in this process, and a lot of smooth edges were ironed out.
But what was the hardware coverage for this? Software? Deviations from the base configuration?
The distros and upstream developers need as much coverage as possible to identify bugs and fix the issues. They also need people who are willing to be patient and file bugs, and help the distros and upstream track down the bugs.
Users may also get burned when filing bugs if they don't see any action on the bugs they are interested in, or have filed. But if they don't file them, the developers won't know about them and can't possibly fix them (unless the developers and/or other testers hit a similar issue).
I ended up installing alsa. It would have been nice if it detected that I wouldn't have the sound driver for alsa installed when it overwrote the modprobe configuration, though.
I should really create an oss4 metapackage to help with configuration. However, I have since done a clean install and pulseaudio and the alsa sound drivers are working reasonably well (a lot better than in previous releases).
There are still some issues (e.g. sound in flash disappears after a while, using Adobe's plugin), but as I understand Ubuntu's configuration of pulse is not correct yet; here's hoping that they finally get it right in Lucid.
There is a balance here.
There are essentially two classes of distribution:
stable (e.g. Debian stable) -- this distribution will not upgrade components until they are settled and working; tends to lag behind on new, shiny features.
latest-and-greatest (e.g. Debian unstable, Ubuntu) -- adds new things and the latest versions of components/drivers without knowing how good they are.
Ubuntu is trying to compete with Windows and Mac, and needs to keep up with evolving features to get the latest functionality (e.g. hot-pluggable input devices in Xorg). It is also focused on meeting requirements of users, so has the option for binary drivers (nvidia, some wifi) so that things can (or should) work.
Exactly. I have been using Karmic from a clean install without major issues, but upgrade killed my sound. Note that I did have OSS4 configured on 9.04 before the upgrade.
One thing I have noticed (and I haven't used other distros to see if this is a common phenomena) is that upgrading Ubuntu is temperamental when it comes to non-standard configurations/customisations (e.g. removing pulseaudio or totem).
Windows: Don't adopt until Service Pack 1
Ubuntu: Don't adopt until 1-2 months after release
Microsoft, Canonical, et. al. are in an interesting position. You need a large number of testers running on a wide variety of hardware (intel, nvidia, ati, ...) with a wide variety of configurations (gnome, KDE, modified sound configurations, heavily modified/customised), needs and requirements (a DJ/sound studio will have different requirements to someone who just wants email). If people don't use the OS because it is 'buggy', the bugs are not found and no fixes are released. Hopefully, they will get good coverage over the alpha and beta phases, then a wider adoption with the enthusiasts and early adopters, followed later by your "average" "novice" user.
Plus, Karmic and other Linux distros are not equivalent to Windows; they are equivalent to Windows, Office, Photoshop and a whole host of other applications, all updated and packaged every 6 months (for karmic, main is about 6.7 GiB and universe is in the 25 GiB mark, not sure about multiverse). That's a lot of stuff to ensure is working on everyones machines.
So the C# list is a std::vector> instead of a std::list -- try comparing like-for-like data structures.
Which is no different to a half-way decent optimiser.
Writing to a pointer does not discard a lot of variables cached in registers; it should discard one if there are no registers free. If it does, the compiler has a very poor registry allocator. It should use one register for a standard loop over contiguous memory for the pointer and continually update that one.
A JIT compiler does have the advantage of being able to do on-the-fly optimisations like you suggest, but the more complex the optimisations, the longer the time needed to generate the jitted code.
The reason iterating through a list is likely to be slower than walking through contiguous memory is that you are likely to require more memory reads with list traversal. Contiguous memory access will keep the data in level 2 cache longer, so the CPU will not be waiting to get the memory.
There are techniques for lists like using a pool allocator -- that is, allocating list items in blocks of n, so you are more likely to keep them in the L2 cache. That is: it all depends on how the data structures are implemented.
Comparing two different data structures is sure to give different performance results. It's like saying a C++ std::vector is faster than a C++ std::list. Also, you need to specify the operations you are performing on the data structures, as they have different performance characteristics (e.g. list has O(1) add and remove, but O(n) random access; vector has O(1) access and but does not guarantee O(1) add/remove (as it may need to allocate memory to expand)).
And also most half-way decent list implementations will block-allocate list nodes in contiguous memory, reusing nodes as they are released. This should yield better allocation performance.
Be aware of the performance characteristics of data structures and algorithms, choosing the best one for the task being performed; perform profiling to see where the slowdowns are and use that to back up data structure/algorithm choices.
I also like how the current model fits nicely with spdf orbital groupings.
Yes - what they are doing is more Chrome-like. They are aiming to reduce visual and UI clutter/redundancy and add better support for things like Vista Glass.
If Firefox were to ribbonise, it would look something like this: http://www.raizlabs.com/blog/uploaded_images/firefox_6-793488.JPG (yuck!)
Since it is the "Office UI", they are referring to other office suites (i.e. OpenOffice, KOffice and friends) as "clone products". See the "Can any applications use the license?" section of the link you referenced.
Not sure whether creating wrappable toolbars grouped in tabs that can resize and show/hide text labels depending on size is worthy of protecting, or even that innovative. The application of the UI may have been new, and the research done well, but the UI itself is really an evolution of toolbars, tabs and tabbed dialogs.
Yes. And TFA is taken initially from https://wiki.mozilla.org/Firefox/Sprints/Windows_Theme_Revamp/Direction_and_Feedback, which is discussing the direction of *applications* written for Vista and Windows 7 that don't use the menubar, but use a contextual strip (Windows Explorer) or Office Ribbon (Paint and Wordpad). That paragraph is about the rationale for not showing the menubar on Vista and later in Firefox, not on adding a ribbon to Firefox (it is under a Hiding of the Menubar section).
It seems as though a blogger misread this paragraph, and everyone on the interweb has been taking this as fact, without actually RTFOA (Reading The Friendly Original Article).
From the pcpro article referenced in the /. summary:
"Starting with Vista, and continuing with Windows 7, the menu bar is going away," notes Mozilla in its plans for revamping the Firefox user interface. "[It will] be replaced with things like the Windows Explorer contextual strip, or the Office Ribbon, [which is] now in Paint and WordPad, too."
From the Mozilla page:
"Starting with Vista, and continuing with Windows 7, the menubar is going away. To be replaced with things like the Windows Explorer contextual strip, or the Office Ribbon(now in Paint and Wordpad too). Many apps still retain the menubar as an option to be pinned or to be shown briefly by holding the Alt key."
Note that here they are talking about Vista and Windows 7, not Firefox (and also note the "Many apps ..." bit in the last sentence).
Translation: crippled to work differently to BIND8 (causing issues when communicating with a real BIND8 server), or providing additional features that only work with Microsoft DNS.
The point is the implication. They say that freeware implementations of DNS are not malware, but are just as bad. They are using freeware to refer to the open source implementations (are there any freeware implementations of DNS - i.e. ones that you don't have to pay for and don't come with the source code) -- see also how they say that theirs is better because you cannot see the source code.
How much of the problems are due to Ubuntu using bleeding-edge packages?
How much of the problems are due to interfacing with binary-only drivers (e.g. nVidia drivers, which are needed for acceleration and 3d graphics; yes, there is nouveau, but that is still a work-in-progress)?
And Greg KH has put in a lot of effort getting the Hyper-V drivers into shape.
Now, lets say that there is some code that the maintainer does not understand. Because they don't have access to the hardware, software (in the case of virtualisation platforms) or technical specs, what should they do?
And yes, it is the *community* (of which Microsoft has become a part). If Microsoft has a vested interest in these drivers, they need to help the community (it is, after all a two-way process).
Don't you mean;
Microsoft: Here is a binary Linux kernel module for Hyper-V. .32 drivers.]
Someone: But wait... it is using GPL-only kernel interfaces.
Microsoft: Oh, er...
Grek KH: You need to release the code to be compliant.
Microsoft: Here you go. Here is the code for the driver.
Microsoft PR: Microsoft has generously donated a lot of code to the Linux kernel under the GPL license. We did this because it is the right thing to do.
[Slashdot story #1]
[Greg KH spends a lot of time thanklessly getting the code to the point where it is ready for inclusion with the kernel.]
[Greg KH sends e-mail to Microsoft asking for help.]
[Greg KH posts a summary for the upcoming
[Slashdot story #2]
Microsoft: Ok, we'll help maintain it.
Microsoft PR: Microsoft is actively involved with the Linux community.
Yes. That was one thing that I immediately noticed -- they have a scrollbar on a menu (that is not even taking up the whole screen)?!
Some other things I noticed:
1/ In the readability section, the apple one is nice and clear (there is a single product, single focus to the page); the microsoft one -- wtf is the page telling me? That I can get a decent laptop for under $1700?! It also has prominent placement for things like IE8, "Windows Live Mail" and "Windows Live" and other Microsoft products.
Another example: go to the Windows 7 site and you get information about what is new about Windows 7. I personally think that this is designed well (it is on point like the apple one is).
Now go to the Vista site and the main content contains an advertisement for buying a new PC now to upgrade to Windows 7 later, with "Thinking about Windows 7?" below. In fact the main bulk of this page is about Windows 7, or buying a laptop that you can upgrade to Windows 7. WTF does this have to do with Vista?
NOTE: Both of these pages prompt you to "Upgrade your internet experience" if not using IE8 and Silverlight. They also have too many link toolbars and search at the top that eats into the vertical space (especially if viewing it on a widescreen monitor/laptop).
2/ The content below the news section on the microsoft page changes when you mouse over it. This causes the content to be refreshed (resulting in the page height changing in some cases, which can cause the page to jump around, and meaning that it is not easy to link to that information or copy sections of it).
That really annoys me too. Plus, searching for Win32 APIs (on the MSDN search -- which uses whatever the Microsoft search engine is called today -- Live, Bing, ...) either gives you results for .NET, or takes you to the WinCE/Mobile API documentation!
Google is the best way as it does not have the broken links (from my observations) and gets you to the right platform.
Plus, I've learned not to rely on what MSDN says -- they even get the API wrong (e.g. it says that the ICreateErrorInfo Set methods take LPCOLESTR when they actually take LPOLESTR arguments).
My Linux computer is infested with zombies: http://www.popcap.com/games/pvz/?icid=plantsvszombies_HP_DL_3_8_19_08_en (Plants vs Zombies game via Wine) ^_^
They have it: Ubuntu logo > Add/Remove.
The applications are there; no need to compile code or anything crazy to install them. Same with other distributions.
What *is* needed is a better way to make this visible to people. Also, add user reviews / ratings / rankings (with some moderation or other system to remove spam and unrelated comments). Add a way to see most popular (today / this week / this month / all time; by download / usage / rating) and "recommended" applications.
Group games by sub-category, so you can see card games, first-person shooters, real-time strategy, etc.
Include screenshots, so people can see what they are like.
Provide links to bug reports (for known/reported) issues as well as forums for help for both Ubuntu (or your favorite distribution of choive) and the application.
Provide "if you like this, people recommend ..." and "people using this are also using ..."
Provide a way for externally supported applications / versions (e.g. Firefox 3.5/Minefield and Songbird). This could help seeing what applications are popular for future distribution releases.
What would be good is if this were done across all distributions. That way you get a unified experience and can help you in distribution choices (which distribution has better KDE4 support?).
... because it updates *everything* (the operating system and all installed applications that come from the distribution).
And the "cannot tell what updates are required and which are optional" comment in Linux is ridiculous. In the update manager on Ubuntu (checked on 9.04), it clearly shows updates with "Important security updates", "Recommended updates" and "Other updates" listed, with a description of the changes.
And people think that by moving to a completely different graphics server is going to fix this?
I use both flac and ogg. I also only use devices that support these formats - currently the Cowon S9 (a very nice device). Cowon are very good at supporting these formats and have for several years.