-- Improve P2P with P4P. Learn more! [pandonetworks.com] I like how their charts start at 400 peers. I can't remember the last time I connected to a torrent that had over 400 peers. Where's the rest of the chart? You know, the part before it plateaus.
This study doesn't show anything but network quality. Furthermore, since so many networks have peering agreements with each other and your data flies around between them readily, it barely judges network quality. What I don't get is I thought the RST packets in Comcast's case were generated by Comcast grabbing the list of peers from you during your communications with the tracker, then sending RST packets to each of the peers you're trying to connect to, aborting your connections with them. The goal being to prevent you from seeding.
So isn't it actually more important to know, for each user, how many RST packets are forged as coming from their IP and sent to other users, not how many are received by each user? That would mean the results of this Vuze sample would be RST packets potentially generated by other ISPs, right?
Technically it doesn't matter where it gets up converted but the general thinking is that up conversion in the player is of a higher quality since it is performed before the signal is compressed to fit over the video cables. What kind of compression is performed to make the signal "fit over the video cables"? That doesn't sound right to me.
Wikipedia has a slightly better explanation:
This often negates the advantages of scaling from within the DVD player, which for technical reasons can be superior to the TVs internal scaler. The DVD player has access to MPEG metadata for each frame of video (interlaced/progressive, motion vectors, etc.) which allow a better deinterlacer and chroma upsampler (these steps are part of the scaling process) than the TV which no longer receives this information. Closer, but not quite right.
How about this? The DVD player has access to the MPEG files, which allows it to look ahead in the stream while performing deinterlacing and upscaling. HDTVs don't have that luxury, and can only use the stream that they are fed (current frame and previous frames, if cached). Much better.
Really, I make heavy use of the bandwidth given me (routine full load) and I've never received any of these notices, any sort of throttling or anything else. I thought the same thing about these articles until Comcast started throttling in my area. Enjoy it while it lasts.
3. The output was a.TS file, and I had a hard time finding free programs that could convert it to standard mpg files. That's a MPEG-2 Transport Stream. Assuming the codec used is also MPEG-2*, you can use Project-X to convert TS to "M2P", aka MPEG-2 Program Stream files, aka your "standard MPEG-2 files." In addition to trans-muxing to a new container, it will also check for timestamp and other errors, join multiple TS files, and a few other things. Free, GPL, Java, cross-platform.
*Unlike some H.264 TS broadcasts in Europe. Those are a bitch to deal with.
Notice they have a "Task Force" for testing Microsoft, but no such group for Firefox, Opera, Safari, etc. Not that surprising, really. There are entire websites devoted to helping web designers hack around IE bugs. If only a single browser could pass Acid2 and Acid3, ideally that browser would be IE. It's used by the most people, so you must design around its flaws. Not to mention, if that were to happen, Firefox and Opera would do everything possible to catch up immediately. Then we wouldn't have to hack around any browser's flaws.
I can't be alone in this. Surely others have this same cognitive disability. I do, in certain specific-purpose MS Word toolbars and many third-party apps. But obviously, I don't have any problems with the forward/back, reload, stop buttons in my browser. Why? Because the icons only have meaning to you after you've used them for some time.
The lesson is that there should only be toolbars for commands you will use enough to start remembering them. That, and the icons should be simple enough to recognize immediately, unlike most third-party apps' specific icons.
I've always thought that brake lights should be a meter from yellow to red, maybe 10 shades, or at least 5. If you see yellow ahead of you, tap the brakes. If you see red, slam them.
People can't seem understand that a small sports car is likely to have a manual transmission, and that I'll be using that instead of braking. Good point. So the brake meter should light up according to all deceleration, not just braking.
They probably use only the basic "blend" form of deinterlacing, rather than checking whether it is best to use IVTC for "Film" (typical DVD) or one of the fancier deinterlacing methods like TomsMoComp for "NTSC" video (typical camcorder).
It's probably best to deinterlace video yourself prior to uploading to YouTube, if you care about quality.
But it looks like the Skateboarding Dog video is pretty poor quality to begin with. Big blurred deinterlaced frames are still blurred deinterlaced frames, regardless of the number of bits you throw at it.
Google style clustering, where you know some of your hardware will fail from time to time and you're just OK with that, is the first promising alternative to mainframe uptimes since the days of VMS clusters. Why should Google care? They can't provide "wrong" search results because of failure. Only out-of-date or not-so-great search results.
Their internal "bigtable" distributed database sounds like it needs better accuracy, but not their actual product.
First off, it let me see what was playing in iTunes without leaving WoW. Yeah, the iTunes G15 plugin is great. No need to switch from any app to change your song with the display and play/pause/next buttons right there on the keyboard. I also find the network monitor plugin handy. It shows your upload/download rates, so you can see when you're maxed out or disconnected.
And the TeamSpeak plugin, to see who is talking to you while in fullscreen games without having to run an "overlay" program. I could see the value in having it on the screen if you need it all the time, but for stuff that's more likely to get in the way most of the time, the extra LCD screen is really helpful.
In fact, I doubt it will ever be a real product. I think it probably will be, at some point. If you look at any media app that sits on top of the filesystem--iTunes, Picasa, etc.--they all share a lot of the same metadata. Having that metadata be part of the filesystem, fully indexed and searchable, makes a whole lot of sense. It can be a separate layer that is independent of the underlying filesystem, as long as it has the right update hooks and every major app uses it. Whether it is MS, or Apple, or Google, or some other 3rd party that provides it, it will happen eventually.
Filenames, extensions, and several incompatible metadata tagging systems is a pretty poor way of abstracting content that is really just Artist, Album, Track, Title, Photographer, Orientation, Resolution, Website, etc. Metadata that is not all hierarchical like a filesystem. Filenames can stay for uniquely identifying a file, but the user shouldn't actually care most of the time what the filenames are, or how to force that information into a hierarchy.
Torrents do that anyway. That is the reason why comcast have to beat them on the head. Each segment in the download is small enough to fit its "booster" criteria. No, that's not right.
PowerBoost only accelerates the connection if the average speed you've been getting over the past 30 seconds* is less than the speed you are rated at/paid for. So if you have a 6 Mbps connection, that's 768 KB/s max. PowerBoost will raise that to up to 2 MB/s for a little less than 15 seconds, making your average for the past 30 seconds equal to 768 KB/s. After that, no matter how many new connections you open, your connection stays at 768 KB/s. But if your connection gets interrupted/throttled for a few seconds, you may get another boost after it resumes, until you are back to 768 KB/s 30 second average again.
*it may be slightly more/less than a 30 second average. Boosts seem to last about 10-15 seconds, which would make sense with that number.
Up until the last month or two, I've had Comcast with no BT filtering and everything was running fine. Up to 400KB/s down on good torrents. Poor upload, but exactly what the agreement stated at 6Mb/384Kb. Kept the upload from being saturated at 48 KB/s by capping it at 35-40 KB/s and web browsing still worked fine with it. Pinged at 35-50 ms while running BT.
The past couple months, web browsing is unbearable while running BT with Comcast. As soon as I start it up, even at 15 KB/s upload, websites take 5-10 seconds to start displaying. Yet I still ping comcast.net and google.com at 35 ms. Strange. BT seems much slower at 100KB down max and sometimes dropping to 15 KB/s up. Same thing, every time. Completely different situation than before.
So, is this what their filtering looks like? Does it affect the whole connection, or just the bittorrent connections? I don't know much about network tools other than ping. Anyone got any pointers on how to check for the RST packets or whatever they're doing?
Anyone know of any user friendly, non-web, multi-platform tagging software? ACDSee is by far the most popular. Windows only. I could have sworn it ran on Mac. Guess not. Contrary to its "Pro" name, it does not deal very well with large numbers of images. It's also fairly easy to corrupt its internal FoxPro database.
iMatch is another Windows alternative, for the database-loving geeky types. Not very user friendly, but full-featured metadata capabilities. Again Windows only.
I haven't found one that meets my needs all that well with adequate performance and open-ness to the software and storage format, so I'm working on a Python/wxPython/Sqlite alternative to ACDSee. It will be cross-platform, Windows, Mac, and Linux. It currently has a fast image viewer, a Sqlite backend, and can import an ACD database, but I'm still working on some of the tagging, rating, and EXIF/IPTC features. It will be somewhat similar to ACDSee and Cornice (another wxPython app), but with tagging features intact, and imho a better looking, more native theme. It will likely be dual licensed. I'll probably have a beta within about a month.
So, long story short, there is no great cross-platform image tagging system right at this moment.;-)
Python could be a much higher performance language. Take a look at the Shed Skin implementation. One guy is trying to implement a hard-code compiler for Python that does type inference to determine types at compile time whenever possible. That yields a 10x-30x speedup. If you have rackmount servers running Python, that's a big win - one rack instead of ten. Or the more popular Psyco, a JIT C-compiler for Python, which boasts "2-100x speedups (typically 4x)", without modifying your code.
Python has some optimization gotchas that really should be fixed. The big problem is "undetectable dynamism", or "if the only tool you have is a dictionary, everything looks like a hash". Well, it sounds like that has issue has mostly been solved. PyPy uses "RPython", a restricted subset of CPython. It "compiles" Python code to C, LLVM,.Net CLI, and Java JVM. It might be interesting to note that the author of Pysco has joined the PyPy project and focused all his optimization efforts on that, as a more generalized implementation of Psyco's goals.
The reason is obvious. Just ask the Fortran guys why they dropped semantic whitespace in the switch from fortran-77 to fortran-90.
Whitespace shouldn't have semantic meaning because then you're stuck with semantic whitespace that might not be the easiest to read. For example, The following is MATLAB code specifying a 10x10 matrix that could be used for future calculations. Since this discussion is about Python, your example doesn't have any relevance. The only significant whitespace in Python is relative indent level, which is tokenized by the lexer in much the same way as curly braces are in other languages--to delineate blocks. Spaces within arrays/lists are not significant, except in string literals of course.
4. Unicode will become the standard throughout Python. This is probably the biggest change. Whereas before the String datatype was synonymous with a byte array (and used for both), it no longer will be. This entails some additional changes in your code to specify the encoding of a String, which will now be represented as Unicode. A new datatype, called "Bytes" (IIRC) can replace the former byte array use of Strings.
5. Many modules in the Standard Library which have been marked deprecated for several versions now (and raised a DeprecationWarning exception) will now actually be removed. Several others with old naming conventions, ambiguous names, or dual Python/C implementations will be renamed, with an appropriate backward-compatible stub that issues the same DeprecationWarning exception, but still works with the old name. A running tally of modules to be removed/renamed is in PEP 3108.
The bottom line is that they are not just making backward-incompatible changes for the sake of changing things. They are specifically trying to avoid "becoming the next Perl 6" to paraphrase Guido from one of his Google Tech Talks. They are just taking this one chance to remove longstanding "warts" from the language in one fell swoop.
Almost everything will be covered by the 2to3 converter. Where required changes are ambiguous, 2to3 will issue a warning and request you fix it manually. I'd rather they fix the problems with the language now before they get any more entrenched. But then, I'm not a maintainer of "legacy" Python code. I am looking forward to the Generic Functions and Adaptation features (among other things) that are being discussed for Python 3000 though.
I went with the 512mb 8600gt instead of thr 256 version because 109 vs 119 USD is trivial with the value of the dollar these days. I didn't realize they were that cheap now. It would help in this case, with its H.264/VC-1 acceleration. Unfortunately, gaming wise the 8600GT is not much of a worthwhile upgrade from the 7600GT. Still waiting for the new G92 version of the 8800GTS 512MB to come down to $200 before making that leap.
$100 is nothing when you just dropped $1400+150 in taxes... Yeah, anything seems cheap compared to the outrageous prices of buying instead of building...
It only takes about an Athlon 2100+ to do either without dropping frames. Not according to Anandtech. Core 2 Duo E6600 will hit 100% CPU usage and start dropping frames with any GPU earlier than GeForce 8-series.
An Athlon 2100+ wouldn't come close to handling it with an older video card. It probably couldn't even with an 8800 GTX GPU or one of the new G92 cards.
You're looking at more like X2 4800+ (probably even higher) on the AMD side of things...
Google has punished sites for shady search engine optimization, but in those cases the sites had always used on-site techniques which could not have been performed by an outsider. Google does penalize for duplicate content. For example, if you setup your domain to have the same content on http://www.slashdot.org/ and http://slashdot.org/ a mirror rather than a redirect (notice the www. is a redirect here). It also penalizes content such as wiki-type content that gets mirrored in several sites around the web. Some webmasters have studied the effect of someone plagiarizing their content in this way and causing that effect. Though obviously their experiments couldn't have been very controlled.
A Google bomb is when many people link to a page and use the same unfitting link text, and then the target page moves UP in the rankings for that particular search term. I agree with you there. It's the only usage I've ever heard of the term. Such as "miserable failure". The first hit for that search used to be Bush's biography at whitehouse.gov, until the articles about the phenomenon itself pushed it down. Google likely fine-tuned their algorithm sometime along the way as well.
I'm no apologist for Comcast but for the last 6 months I've been able to speedtest at 22+ mb down and 2+ mb up. Speedtest is worthless for Comcast because of Powerboost. You need to download/upload for at least 1 minute each to get a reliable average after the much-inflated 10 second or so boost at the beginning.
Usenet, powerful uncapped web servers, and extremely well-seeded torrents are good for testing download. FTP or almost any torrent with peers is good for testing upload. You'll likely see that your real speed that you get in the second minute and every subsequent minute, provided you don't reset the connection or artificially cap the download speed long enough for Powerboost to kick in again, is capped at what you paid for in download speed (probably 6 Mbps) and 384 Kbps upload.
The way powerboost works is if your average D/L speed for the past 30 seconds or so is less than your rated speed, it will more than double the capped speed until that 30 second average is achieved, then lower the cap to exactly what you paid for after your 30-second average meets that cap. It's pretty precise. You're not getting anything beyond what any other Comcast customer gets, no matter what Speedtest leads you to believe.
In it he advocates the benefits of Range Voting as a solution to Arrow's Impossibility Theorem. Ever been to a site that allows people to vote on articles on a scale of 1-10? It rapidly degenerates into everyone either voting 10 or 0, based upon whether they think the article is overrated or underrated. Basically, if you don't vote in a binary fashion like that, your vote doesn't count as much.
Might as well just go with the simpler Approval voting, mentioned in the wikipedia article you linked:
However, approval voting is range voting with only 2 levels (approved (1) and disapproved (0)) and forms of approval voting have been used for example, in Venice in the 13th century. It's simpler, and more effective in my experience.
Slashdot Mirror
Improve P2P with P4P. Learn more! [pandonetworks.com] I like how their charts start at 400 peers. I can't remember the last time I connected to a torrent that had over 400 peers. Where's the rest of the chart? You know, the part before it plateaus.
So isn't it actually more important to know, for each user, how many RST packets are forged as coming from their IP and sent to other users, not how many are received by each user? That would mean the results of this Vuze sample would be RST packets potentially generated by other ISPs, right?
Wikipedia has a slightly better explanation: This often negates the advantages of scaling from within the DVD player, which for technical reasons can be superior to the TVs internal scaler. The DVD player has access to MPEG metadata for each frame of video (interlaced/progressive, motion vectors, etc.) which allow a better deinterlacer and chroma upsampler (these steps are part of the scaling process) than the TV which no longer receives this information. Closer, but not quite right.
How about this? The DVD player has access to the MPEG files, which allows it to look ahead in the stream while performing deinterlacing and upscaling. HDTVs don't have that luxury, and can only use the stream that they are fed (current frame and previous frames, if cached). Much better.
*Unlike some H.264 TS broadcasts in Europe. Those are a bitch to deal with.
The lesson is that there should only be toolbars for commands you will use enough to start remembering them. That, and the icons should be simple enough to recognize immediately, unlike most third-party apps' specific icons.
They probably use only the basic "blend" form of deinterlacing, rather than checking whether it is best to use IVTC for "Film" (typical DVD) or one of the fancier deinterlacing methods like TomsMoComp for "NTSC" video (typical camcorder).
It's probably best to deinterlace video yourself prior to uploading to YouTube, if you care about quality.
I don't see it.
But it looks like the Skateboarding Dog video is pretty poor quality to begin with. Big blurred deinterlaced frames are still blurred deinterlaced frames, regardless of the number of bits you throw at it.
Their internal "bigtable" distributed database sounds like it needs better accuracy, but not their actual product.
And the TeamSpeak plugin, to see who is talking to you while in fullscreen games without having to run an "overlay" program. I could see the value in having it on the screen if you need it all the time, but for stuff that's more likely to get in the way most of the time, the extra LCD screen is really helpful.
In fact, I doubt it will ever be a real product. I think it probably will be, at some point. If you look at any media app that sits on top of the filesystem--iTunes, Picasa, etc.--they all share a lot of the same metadata. Having that metadata be part of the filesystem, fully indexed and searchable, makes a whole lot of sense. It can be a separate layer that is independent of the underlying filesystem, as long as it has the right update hooks and every major app uses it. Whether it is MS, or Apple, or Google, or some other 3rd party that provides it, it will happen eventually.
Filenames, extensions, and several incompatible metadata tagging systems is a pretty poor way of abstracting content that is really just Artist, Album, Track, Title, Photographer, Orientation, Resolution, Website, etc. Metadata that is not all hierarchical like a filesystem. Filenames can stay for uniquely identifying a file, but the user shouldn't actually care most of the time what the filenames are, or how to force that information into a hierarchy.
PowerBoost only accelerates the connection if the average speed you've been getting over the past 30 seconds* is less than the speed you are rated at/paid for. So if you have a 6 Mbps connection, that's 768 KB/s max. PowerBoost will raise that to up to 2 MB/s for a little less than 15 seconds, making your average for the past 30 seconds equal to 768 KB/s. After that, no matter how many new connections you open, your connection stays at 768 KB/s. But if your connection gets interrupted/throttled for a few seconds, you may get another boost after it resumes, until you are back to 768 KB/s 30 second average again.
*it may be slightly more/less than a 30 second average. Boosts seem to last about 10-15 seconds, which would make sense with that number.
Up until the last month or two, I've had Comcast with no BT filtering and everything was running fine. Up to 400KB/s down on good torrents. Poor upload, but exactly what the agreement stated at 6Mb/384Kb. Kept the upload from being saturated at 48 KB/s by capping it at 35-40 KB/s and web browsing still worked fine with it. Pinged at 35-50 ms while running BT.
The past couple months, web browsing is unbearable while running BT with Comcast. As soon as I start it up, even at 15 KB/s upload, websites take 5-10 seconds to start displaying. Yet I still ping comcast.net and google.com at 35 ms. Strange. BT seems much slower at 100KB down max and sometimes dropping to 15 KB/s up. Same thing, every time. Completely different situation than before.
So, is this what their filtering looks like? Does it affect the whole connection, or just the bittorrent connections? I don't know much about network tools other than ping. Anyone got any pointers on how to check for the RST packets or whatever they're doing?
iMatch is another Windows alternative, for the database-loving geeky types. Not very user friendly, but full-featured metadata capabilities. Again Windows only.
I haven't found one that meets my needs all that well with adequate performance and open-ness to the software and storage format, so I'm working on a Python/wxPython/Sqlite alternative to ACDSee. It will be cross-platform, Windows, Mac, and Linux. It currently has a fast image viewer, a Sqlite backend, and can import an ACD database, but I'm still working on some of the tagging, rating, and EXIF/IPTC features. It will be somewhat similar to ACDSee and Cornice (another wxPython app), but with tagging features intact, and imho a better looking, more native theme. It will likely be dual licensed. I'll probably have a beta within about a month.
So, long story short, there is no great cross-platform image tagging system right at this moment.
Whitespace shouldn't have semantic meaning because then you're stuck with semantic whitespace that might not be the easiest to read. For example, The following is MATLAB code specifying a 10x10 matrix that could be used for future calculations. Since this discussion is about Python, your example doesn't have any relevance. The only significant whitespace in Python is relative indent level, which is tokenized by the lexer in much the same way as curly braces are in other languages--to delineate blocks. Spaces within arrays/lists are not significant, except in string literals of course.
To expand on that:
4. Unicode will become the standard throughout Python. This is probably the biggest change. Whereas before the String datatype was synonymous with a byte array (and used for both), it no longer will be. This entails some additional changes in your code to specify the encoding of a String, which will now be represented as Unicode. A new datatype, called "Bytes" (IIRC) can replace the former byte array use of Strings.
5. Many modules in the Standard Library which have been marked deprecated for several versions now (and raised a DeprecationWarning exception) will now actually be removed. Several others with old naming conventions, ambiguous names, or dual Python/C implementations will be renamed, with an appropriate backward-compatible stub that issues the same DeprecationWarning exception, but still works with the old name. A running tally of modules to be removed/renamed is in PEP 3108.
The bottom line is that they are not just making backward-incompatible changes for the sake of changing things. They are specifically trying to avoid "becoming the next Perl 6" to paraphrase Guido from one of his Google Tech Talks. They are just taking this one chance to remove longstanding "warts" from the language in one fell swoop.
Almost everything will be covered by the 2to3 converter. Where required changes are ambiguous, 2to3 will issue a warning and request you fix it manually. I'd rather they fix the problems with the language now before they get any more entrenched. But then, I'm not a maintainer of "legacy" Python code. I am looking forward to the Generic Functions and Adaptation features (among other things) that are being discussed for Python 3000 though.
An Athlon 2100+ wouldn't come close to handling it with an older video card. It probably couldn't even with an 8800 GTX GPU or one of the new G92 cards.
You're looking at more like X2 4800+ (probably even higher) on the AMD side of things...
Usenet, powerful uncapped web servers, and extremely well-seeded torrents are good for testing download. FTP or almost any torrent with peers is good for testing upload. You'll likely see that your real speed that you get in the second minute and every subsequent minute, provided you don't reset the connection or artificially cap the download speed long enough for Powerboost to kick in again, is capped at what you paid for in download speed (probably 6 Mbps) and 384 Kbps upload.
The way powerboost works is if your average D/L speed for the past 30 seconds or so is less than your rated speed, it will more than double the capped speed until that 30 second average is achieved, then lower the cap to exactly what you paid for after your 30-second average meets that cap. It's pretty precise. You're not getting anything beyond what any other Comcast customer gets, no matter what Speedtest leads you to believe.
Might as well just go with the simpler Approval voting, mentioned in the wikipedia article you linked: However, approval voting is range voting with only 2 levels (approved (1) and disapproved (0)) and forms of approval voting have been used for example, in Venice in the 13th century. It's simpler, and more effective in my experience.