1. Microsoft doesn't give a shit about it. Therefore enterprise users won't give a shit.
2. Even if Microsoft does give a shit, neither Apple nor Microsoft will support Ogg Theora. Therefore Linux is SOL again.
3. Apply #1 and 2 to audio standards as well. No common, open, royalty free, pre-installed standard across all platforms == epic fail.
The main power of Flash right now is that once you install the plugin, you might as well forget all that BS about paying for codecs on all three major platforms. It's all in there. It's convenient. It's sufficient.
Um, no. You can't keep someone else's code 100% in your head 100% of the time, even if you're paired with the guy. Heck on anything non-trivial it barely works with the code you've written yourself. By "keeping it in your head" I mean being able to anticipate, with full clarity, the effects of changing a bit of code nominally unrelated to the code where you'd be causing change in behavior. Things like exception handling, exception safety, thread safety, concurrency, performance implications, validation, security, etc.
You simply won't have the same level of grasp on the code written by someone else.
Remember, the best code in the world is 100% written without all of this newfangled "agile" BS. Pick any successful open (or closed) source project and try to find one that uses anything Agile-related (XP, Scrum, PP). You won't find one.
The article is most likely BS, written by someone who doesn't know how the Search industry works. Let me lay out some facts for you so that you see why I think it's BS:
1. Most, if not all algorithms that Live Search / Bing uses are PUBLICLY DISCLOSED in papers published by Microsoft Research, and the corresponding patents that Microsoft holds. You don't need to "identify" anything. And even if you did, the new features introduced by Bing are so superficial that "identifying" similar algorithms would not take Google's engineers and researchers much time.
2. All major search engines monitor each other constantly and they know exactly what the competition's NDCG metrics (normalized cumulative distributed gain - the measure of how relevant the results are) are. As a rule, it's undesirable to crank up the NDCG too much, since doing so reduces the click through rate on ads, so historically, Google has kept their NDCG just a wee bit ahead of Yahoo and Live, and every time the two would update their algorithms, Google would crank it up a notch to stay ahead. To think that they've been sitting on their ass in the past couple of years is stupid.
So at most, I think Google is working on some experimental stuff related to presentation of results, which is where it's currently lacking, in spite of their half assed, hidden-by-default sidebar.
This site basically outputs search results in three columns, with all formatting uniform, all branding removed and columns permuted on every search. You vote for the best results. I found myself unknowingly "voting" for Bing a surprising number of times.
I was about to say this. I've worked with some really good developers and I consider myself "good" as well. We would not do code reviews for _all_ new code. We would ask each other to review particularly hairy pieces, though, which would often result in insightful and non-obvious refactoring suggestions, and rarely in finding bugs. Security related bits would always get a full-team review (including QA). Code reviews were required for all but the most trivial bug fixes, since they're cheaper and you don't always remember what the old code did 100%. Closer to shipping, two reviewers were required. There were very few bugs overall.
I'm pretty confident that I can code faster and with higher quality than pair programmers of virtually any skill level. If they're bad, they'll inevitably churn out bad code (although it will be better than code written by just one bad programmer), and if they're good, pair programming will slow them down and lead to more defects because neither of the programmers will have the whole thing 100% in his head.
They aren't comparable. Cloudbase is a simple SQL layer on top of Hadoop that operates on flat files. HBase is an open source BigTable (i.e. you can't really do SQL with it). Hive is kind of like Cloudbase. In the end all of these systems have a common strength/weakness - Hadoop. The strength is that they scale, if you're willing to pay, the weakness is that their scalability seems to be piss poor on smaller clusters.
In the end we went with bare Hadoop operating directly on LZO compressed log chunks. Log chunks are written to HDFS by a distributed grid of daemons. For ETL and preliminary cleanup, this works fine on 1TB of log data per day. If we had less than that, we'd import directly into Aster DB, which is what we use for DW and ad-hoc data analysis.
I've evaluated Hadoop (and Cloudbase, HBase and a few other things) for transaction log mining purposes and found it to be VERY inefficient. Basically, if your machine has a decent RAID array (by "decent" I mean 500-700MB/sec linear read throughput, and 300-500MB/sec write throughput), you will need 12-15 8 core Hadoop boxes to even come close to a single machine's performance. This, IMO, is fucked up. I expected it would be much more efficient than it is.
Therefore, my conclusion was that Hadoop only makes sense when you can't solve a problem any other way and are prepared to pay through the nose for hundreds or thousands of machines to alleviate its performance shortcomings.
Caveat lector - my biggest Hadoop cluster consisted of 20 8-core nodes, with 32GB RAM per node and GigE interconnect.
"Bolt" has already shown that Lasseter is only a part of Pixar's magic. "Bolt" was a POS Disney movie, just rendered in 3D. As it often happens, the higher ups aren't really worth that much without the right set of grunts in the trenches.
Calling those two boxes "servers" was probably too generous. They just run headless and do compute work pretty much non-stop, that's why I called them "servers". I use CentOS for internet-facing servers.
Ultimately, couldn't see why I'd stick with Fedora anymore. Most of my servers run CentOS (which is a humongous pain in the ass), the new Ubuntu servers are for development only. If Ubuntu does well there, CentOS may have to give way, too.
That's what virtualization is for. I always install a new OS in a VM first. This time around, doing this led me to switching a couple of servers from Fedora to Ubuntu Server.
Seems to me like people in Europe enjoy more freedoms than we do here in the US - the self proclaimed "Land of the Free, Home of the Brave".
That's what you get with a single party system, my friends. And no, this is not a typo - Dems and Repubs are pretty much the same party with minor variations. There's nowhere near the diversity of political opinion in the US as what you'd see in Europe. We need a raving, rabid, card carrying socialists to balance the equation somewhat on this side of the pond. All branches of the government have been licking the Big Business' behind for far too long.
Here are the conditions under which I will agree to pay my money for Hulu:
1. No ads in the paid content. AT ALL. Not now, not ever. 2. Cheap, a-la carte subscriptions for individual shows. If I only need a few shows from Discovery, Nickelodeon and Food Network, I should be able to sign-up for only those shows. 3. Compatibility with an inexpensive hardware device of some sort (Apple TV, Xbox or PS3 will do). 4. Content is served in _at least_ 720p with high encoding quality.
These conditions are not negotiable. If all four are fulfilled, I, for one, will welcome our money charging overlords.
It must be nice for a hardware company to build its own datacenters. They get the hardware at cost, and since they're one of the largest purchasers of components, their cost is probably better than Google's. Which is not to say that XServe comes out being cheaper than Google's bare bone server, but you get my drift. Throw some Cisco and F5 on all this goodness, hook up HVAC and UPS and it's all ready to go.
"Work with Active Directory" is a vague notion in Linux. Technically, you can install Likewise Open and join your Linux machine to a domain within 15 minutes. But then you lose the ability to add your AD-authenticated user to local groups (!). You actually have to edit obscure text files by hand to do that, the usual tools simply do nothing. Editing those files incorrectly can lock you out of your machine. All other pieces of software are NOT GUARANTEED to pick up the config. Want to open an OpenOffice doc from a network share? Tough luck. Want Apache to transparently pick up kerberos? Who told you that should work? Want your Java app open files on a remote machine or otherwise use Kerberos? Tough luck again - you'll have to configure it all over, and over, and over again.
Tried that, too. It's only slightly better than VNC, and you see JPEG compression artifacts on your display. Well, in fact it's a modification of VNC, so it can't be dramatically better.
Two things for you: 1. Try RDP first. I have successfully logged in and worked with a machine halfway across the globe connected with a 36.6kbps modem with RDP. Try that with X. 2. Compression does not help a chatty protocol to become any less chatty and does not remove latency, which ruins everything for chatty protocols. There's a reason why people use VNC on Linux.
And BTW, xmove hasn't been updated since 1997. 12 years. Which means no one is using it.
Have you actually TRIED using X Window, even over broadband? I have. There are several drawbacks to it: 1. It is VERY slow, slower than VNC. It runs like a snail compared to RDP. 2. If you get disconnected, for whatever reason, your apps die and you lose data. In year 2009 this is UNACCEPTABLE. 3. Did I mention it is VERY slow?
Seriously, I've been running Linux as my primary OS for a while now, and my work laptop is joined to Active Directory at work through Likewise Open. Even so, the integration is rudimentary at best, and every piece of software has its own little tweaks and settings. Single sign-on is a PAIN on Linux. Group policies don't exist. Peripheral compatibility is spotty, particularly with scanners. Multi-factor auth is a pain in the ass. Remote desktop (VNC) is really slow compared to RDP which makes VPN-from-home scenario painful.
Those are a few MUST HAVE things that work in Windows out of the box. RedHat should hire a few more engineers and get them cracking on those, before spending a ton of money on lawyers.
I do think that they could have supplied quite a bit on the server side, though. File serving, web serving, document sharing, DB - those things don't need Windows anymore.
USSR was a socialist country, not a communist one. There's a big difference between the two. Communism was promised, eventually, but they never got there. Just because the ruling party was CPSU, doesn't mean they had communism.
Slashdot Mirror
For several reasons:
1. Microsoft doesn't give a shit about it. Therefore enterprise users won't give a shit.
2. Even if Microsoft does give a shit, neither Apple nor Microsoft will support Ogg Theora. Therefore Linux is SOL again.
3. Apply #1 and 2 to audio standards as well. No common, open, royalty free, pre-installed standard across all platforms == epic fail.
The main power of Flash right now is that once you install the plugin, you might as well forget all that BS about paying for codecs on all three major platforms. It's all in there. It's convenient. It's sufficient.
Um, no. You can't keep someone else's code 100% in your head 100% of the time, even if you're paired with the guy. Heck on anything non-trivial it barely works with the code you've written yourself. By "keeping it in your head" I mean being able to anticipate, with full clarity, the effects of changing a bit of code nominally unrelated to the code where you'd be causing change in behavior. Things like exception handling, exception safety, thread safety, concurrency, performance implications, validation, security, etc.
You simply won't have the same level of grasp on the code written by someone else.
Remember, the best code in the world is 100% written without all of this newfangled "agile" BS. Pick any successful open (or closed) source project and try to find one that uses anything Agile-related (XP, Scrum, PP). You won't find one.
The article is most likely BS, written by someone who doesn't know how the Search industry works. Let me lay out some facts for you so that you see why I think it's BS:
1. Most, if not all algorithms that Live Search / Bing uses are PUBLICLY DISCLOSED in papers published by Microsoft Research, and the corresponding patents that Microsoft holds. You don't need to "identify" anything. And even if you did, the new features introduced by Bing are so superficial that "identifying" similar algorithms would not take Google's engineers and researchers much time.
2. All major search engines monitor each other constantly and they know exactly what the competition's NDCG metrics (normalized cumulative distributed gain - the measure of how relevant the results are) are. As a rule, it's undesirable to crank up the NDCG too much, since doing so reduces the click through rate on ads, so historically, Google has kept their NDCG just a wee bit ahead of Yahoo and Live, and every time the two would update their algorithms, Google would crank it up a notch to stay ahead. To think that they've been sitting on their ass in the past couple of years is stupid.
So at most, I think Google is working on some experimental stuff related to presentation of results, which is where it's currently lacking, in spite of their half assed, hidden-by-default sidebar.
Compare them yourself, without branding: http://blindsearch.fejus.com/
This site basically outputs search results in three columns, with all formatting uniform, all branding removed and columns permuted on every search. You vote for the best results. I found myself unknowingly "voting" for Bing a surprising number of times.
I was about to say this. I've worked with some really good developers and I consider myself "good" as well. We would not do code reviews for _all_ new code. We would ask each other to review particularly hairy pieces, though, which would often result in insightful and non-obvious refactoring suggestions, and rarely in finding bugs. Security related bits would always get a full-team review (including QA). Code reviews were required for all but the most trivial bug fixes, since they're cheaper and you don't always remember what the old code did 100%. Closer to shipping, two reviewers were required. There were very few bugs overall.
I'm pretty confident that I can code faster and with higher quality than pair programmers of virtually any skill level. If they're bad, they'll inevitably churn out bad code (although it will be better than code written by just one bad programmer), and if they're good, pair programming will slow them down and lead to more defects because neither of the programmers will have the whole thing 100% in his head.
They aren't comparable. Cloudbase is a simple SQL layer on top of Hadoop that operates on flat files. HBase is an open source BigTable (i.e. you can't really do SQL with it). Hive is kind of like Cloudbase. In the end all of these systems have a common strength/weakness - Hadoop. The strength is that they scale, if you're willing to pay, the weakness is that their scalability seems to be piss poor on smaller clusters.
In the end we went with bare Hadoop operating directly on LZO compressed log chunks. Log chunks are written to HDFS by a distributed grid of daemons. For ETL and preliminary cleanup, this works fine on 1TB of log data per day. If we had less than that, we'd import directly into Aster DB, which is what we use for DW and ad-hoc data analysis.
I've evaluated Hadoop (and Cloudbase, HBase and a few other things) for transaction log mining purposes and found it to be VERY inefficient. Basically, if your machine has a decent RAID array (by "decent" I mean 500-700MB/sec linear read throughput, and 300-500MB/sec write throughput), you will need 12-15 8 core Hadoop boxes to even come close to a single machine's performance. This, IMO, is fucked up. I expected it would be much more efficient than it is.
Therefore, my conclusion was that Hadoop only makes sense when you can't solve a problem any other way and are prepared to pay through the nose for hundreds or thousands of machines to alleviate its performance shortcomings.
Caveat lector - my biggest Hadoop cluster consisted of 20 8-core nodes, with 32GB RAM per node and GigE interconnect.
"Bolt" has already shown that Lasseter is only a part of Pixar's magic. "Bolt" was a POS Disney movie, just rendered in 3D. As it often happens, the higher ups aren't really worth that much without the right set of grunts in the trenches.
Calling those two boxes "servers" was probably too generous. They just run headless and do compute work pretty much non-stop, that's why I called them "servers". I use CentOS for internet-facing servers.
Throw more RAM in. It's so cheap these days, I've maxed out the RAM on all of my desktop machines.
You don't need it to be terribly fast if you just want to evaluate.
Ultimately, couldn't see why I'd stick with Fedora anymore. Most of my servers run CentOS (which is a humongous pain in the ass), the new Ubuntu servers are for development only. If Ubuntu does well there, CentOS may have to give way, too.
That's what virtualization is for. I always install a new OS in a VM first. This time around, doing this led me to switching a couple of servers from Fedora to Ubuntu Server.
Seems to me like people in Europe enjoy more freedoms than we do here in the US - the self proclaimed "Land of the Free, Home of the Brave".
That's what you get with a single party system, my friends. And no, this is not a typo - Dems and Repubs are pretty much the same party with minor variations. There's nowhere near the diversity of political opinion in the US as what you'd see in Europe. We need a raving, rabid, card carrying socialists to balance the equation somewhat on this side of the pond. All branches of the government have been licking the Big Business' behind for far too long.
Here are the conditions under which I will agree to pay my money for Hulu:
1. No ads in the paid content. AT ALL. Not now, not ever.
2. Cheap, a-la carte subscriptions for individual shows. If I only need a few shows from Discovery, Nickelodeon and Food Network, I should be able to sign-up for only those shows.
3. Compatibility with an inexpensive hardware device of some sort (Apple TV, Xbox or PS3 will do).
4. Content is served in _at least_ 720p with high encoding quality.
These conditions are not negotiable. If all four are fulfilled, I, for one, will welcome our money charging overlords.
And aside from the expanded snippet feature (which is pretty clever) I don't see anything new or exciting. WTF?
Wave is basically a private forum with versioning and file sharing. I don't see how it's much more "conceptually complex" than Slashdot.
Maybe Mono isn't that "bad" for Open Source after all. A little competition will put a damper on stupid shakedown attempts like this one.
It must be nice for a hardware company to build its own datacenters. They get the hardware at cost, and since they're one of the largest purchasers of components, their cost is probably better than Google's. Which is not to say that XServe comes out being cheaper than Google's bare bone server, but you get my drift. Throw some Cisco and F5 on all this goodness, hook up HVAC and UPS and it's all ready to go.
"Work with Active Directory" is a vague notion in Linux. Technically, you can install Likewise Open and join your Linux machine to a domain within 15 minutes. But then you lose the ability to add your AD-authenticated user to local groups (!). You actually have to edit obscure text files by hand to do that, the usual tools simply do nothing. Editing those files incorrectly can lock you out of your machine. All other pieces of software are NOT GUARANTEED to pick up the config. Want to open an OpenOffice doc from a network share? Tough luck. Want Apache to transparently pick up kerberos? Who told you that should work? Want your Java app open files on a remote machine or otherwise use Kerberos? Tough luck again - you'll have to configure it all over, and over, and over again.
WTF is this? Kindergarten?
And I haven't paid a cent for Linux in a long time. Quality will have to improve quite a bit before I consider spending cash on it.
Tried that, too. It's only slightly better than VNC, and you see JPEG compression artifacts on your display. Well, in fact it's a modification of VNC, so it can't be dramatically better.
Two things for you:
1. Try RDP first. I have successfully logged in and worked with a machine halfway across the globe connected with a 36.6kbps modem with RDP. Try that with X.
2. Compression does not help a chatty protocol to become any less chatty and does not remove latency, which ruins everything for chatty protocols. There's a reason why people use VNC on Linux.
And BTW, xmove hasn't been updated since 1997. 12 years. Which means no one is using it.
Have you actually TRIED using X Window, even over broadband? I have. There are several drawbacks to it:
1. It is VERY slow, slower than VNC. It runs like a snail compared to RDP.
2. If you get disconnected, for whatever reason, your apps die and you lose data. In year 2009 this is UNACCEPTABLE.
3. Did I mention it is VERY slow?
Seriously, I've been running Linux as my primary OS for a while now, and my work laptop is joined to Active Directory at work through Likewise Open. Even so, the integration is rudimentary at best, and every piece of software has its own little tweaks and settings. Single sign-on is a PAIN on Linux. Group policies don't exist. Peripheral compatibility is spotty, particularly with scanners. Multi-factor auth is a pain in the ass. Remote desktop (VNC) is really slow compared to RDP which makes VPN-from-home scenario painful.
Those are a few MUST HAVE things that work in Windows out of the box. RedHat should hire a few more engineers and get them cracking on those, before spending a ton of money on lawyers.
I do think that they could have supplied quite a bit on the server side, though. File serving, web serving, document sharing, DB - those things don't need Windows anymore.
USSR was a socialist country, not a communist one. There's a big difference between the two. Communism was promised, eventually, but they never got there. Just because the ruling party was CPSU, doesn't mean they had communism.