Wikipedia:Reference desk/Archives/Computing/2010 January 14
Computing desk | ||
---|---|---|
< January 13 | << Dec | January | Feb >> | January 15 > |
Welcome to the Wikipedia Computing Reference Desk Archives |
---|
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages. |
January 14
[edit]System Restore to get rid of malware
[edit]My Windows XP Home Edition system was recently infected by the malware which replaces the background wallpaper with "Your System Is Infected!" and other verbiage, including the mispelled word "recommeded". It pops up messages asking you to get security software, and disables system programs like Task Manager, Command Prompt, etc. Anyway, I think I eventually fixed the system by a combination of Malwarebytes' Anti-Malware, and then when it didn't quite do the trick (maybe a little unfair to MBAM since it hadn't been updated in a couple of months), a System Restore (after booting into Safe Mode and logging into the Adminstrator account). All this was new to me; I'm not a Windows expert and hadn't even heard of System Restore before! I was really glad to find it.
However, I've been googling this kind of malware, and I see a lot of recommendations (including what seems to me to be overkill, like using 5 different antivirus programs, or using powerful programs that could potential screw up your system worse if you aren't careful) and no mention of using the System Restore feature, so I'm wondering if I'm missing something here. I realize that going back to a prior checkpoint may roll back your software installations or upgrades that were made since that time, but if you aren't worried about that, is there any problem with System Restore? It was very easy and convenient to use, so I'm wondering if there's a reason why it isn't one of the fist things to try. TresÁrboles (talk) 01:48, 14 January 2010 (UTC) Argh, I misspelled "misspelled" ... TresÁrboles (talk) 01:50, 14 January 2010 (UTC)
- Quite often System Restore won't work; some viruses now infect your restore files as well. It may or may not work. HalfShadow 03:16, 14 January 2010 (UTC)
- I assume you are not talking about restoring from a Restore Disc, that should be fool proof. (Unless the disc, supplied with the system is infected!). It is possible that the on HDD restore files could be infected. But if you restore to a point before the MalWare got in, it should work. See System Restore#Limitations & complications. If you have all your data files backed up then a HDD format and re-install of your O/S (operating system - XP), from the original DVD will almost certainly get rid of any malicious software on your C: or system drive. Though you will have to install any Service packs, patches, driver updates etc. I don't think that System Restore would help in tha case of the infection you had. (I had it too! and also used Malwarebytes' Anti-Malware) It's a bit more for when you have a driver that screws up your sytem or a HDD crash damages some O/S files.
- Good software to have is Ghost, which makes an 'image' file of your HDD or selected partition. If you get you PC clean, you make an image of it, and if it gets infected again, you write the image back to your PC and it will take it back to the exact state it in was when you made the image file. It is a good idea to keep data files separate from the system files (on another HDDentirely) as it's safer, IMHO, you can more easily replace the System files without worrying about your data.
- The idea of using "5 different antivirus programs" is likely because each has it strengths, and may pick up malware that the others miss. I use only AVG Anti-Virus Free, (which is FREE) and find it works quite well. Also ensure that your O/S updates are current, that will help keep malware out. (We ALL make speeling errors!)--220.101.28.25 (talk) 05:15, 14 January 2010 (UTC)
- Hi, why do you say that System Restore wouldn't work? It certainly looks like it works; the background image (wallpaper) is back to normal, I'm not getting messages about infections, I can start and use utilities like Task Manager and Command Prompt now. One thing that I didn't mention above but which also works now is I can update MBAM without a "732 error". Also, I guess a symptom some people had, but I never experienced because I stopped using my browser (Firefox) until the problem was fixed, was that when browsing, they would get redirected. Well that isn't happening either. TresÁrboles (talk) 01:53, 15 January 2010 (UTC)
- I assume you are not talking about restoring from a Restore Disc, that should be fool proof. (Unless the disc, supplied with the system is infected!). It is possible that the on HDD restore files could be infected. But if you restore to a point before the MalWare got in, it should work. See System Restore#Limitations & complications. If you have all your data files backed up then a HDD format and re-install of your O/S (operating system - XP), from the original DVD will almost certainly get rid of any malicious software on your C: or system drive. Though you will have to install any Service packs, patches, driver updates etc. I don't think that System Restore would help in tha case of the infection you had. (I had it too! and also used Malwarebytes' Anti-Malware) It's a bit more for when you have a driver that screws up your sytem or a HDD crash damages some O/S files.
- Running multiple versions of antivirus software is a bad idea because it will hammer your performance, badly; every time you visit a website and your browser downloads the 70 files necessary to view that page, the 5 antivirus programs will compete to compare each file to all the virus signatures in its database ... it'll degrade your system performance a lot. My standard advice is to run 1 antivirus program and keep it updated; back up your data regularly onto an external USB hard disk; and set up your computer so there is one account with administrator privileges, which you use only to install trusted software; and one account with no administrator privileges, which is the account you will use for your ordinary computer use. This will greatly reduce the probability that malware will be able to compromise your system. Comet Tuttle (talk) 07:01, 14 January 2010 (UTC)
- I think those articles/blogs I saw weren't talking about the always-on resident virus scanners that look at things while you work, but manually-running programs... but I see several users are debating this below... TresÁrboles (talk) 01:53, 15 January 2010 (UTC)
- Running multiple versions of antivirus software is a bad idea because it will hammer your performance, badly; every time you visit a website and your browser downloads the 70 files necessary to view that page, the 5 antivirus programs will compete to compare each file to all the virus signatures in its database ... it'll degrade your system performance a lot. My standard advice is to run 1 antivirus program and keep it updated; back up your data regularly onto an external USB hard disk; and set up your computer so there is one account with administrator privileges, which you use only to install trusted software; and one account with no administrator privileges, which is the account you will use for your ordinary computer use. This will greatly reduce the probability that malware will be able to compromise your system. Comet Tuttle (talk) 07:01, 14 January 2010 (UTC)
- While it is sensible to only have one "on all the time" anti-virus program, you can use as many anti-malware scanners as you like. They do not cover the same ground so it is sensible to scan with more than one. In addition to Avast! on-all-the-time, I use Malwarebytes, Superantispyware, AdvancedSystemCare, Spybot Search & Destroy, and SpywareBlaster which works in a different way. Windows Defender is something I'm going to add (Edit - noo I'm not, its not merely a scanner). I also frequently use Ccleaner to remove the junk before scanning. 89.243.186.173 (talk) 10:34, 14 January 2010 (UTC)
- Silly me again. Should have distinguished between 'Scan & Clean' and 'Resident' Antivirus software. Definately not a good idea to have 5 resident programs simultaneously performing the same job(antivirus). Comet Tuttles' advice is very good, especially the bit about accounts, often forgotten about (& I'm talkin about me!) --220.101.28.25 (talk) 12:15, 14 January 2010 (UTC)
- You're wrong, 89! They will interfere with each other. One will prevent the other from disinfecting critical system files. You can install more than one anti-malware program. But you have to make sure that they're not running all the time, because they tend to keep an eye on the same files for modifications. One way to protect your computer without slowing it down is to create a non-administrative account, as Comet Turtle suggested. Another solution is to secure your browser. I use Internet Explorer with Privoxy. Privoxy blocks ads (which often hijack your computer) and filters many exploits. If you use Firefox, then install NoScript and Adblock Plus. If you download any files with a .exe, .com, .bat, .vbs, or .js extension from the internet, then scan them with an anti-virus program. Also disable VBA inside Microsoft Office and JavaScript inside Adobe Reader. Then, you'll be secure from 99.99% of viruses without slowing your computer down or having to deal with some electronic nanny that always gets in the way of things.
- On another note, I'd like to add that since your system restore points may now be infected, you should delete them.--Drknkn (talk) 13:16, 14 January 2010 (UTC)
- I have now installed NoScript. I have to point out that I didn't download any unknown .exe file or the like and run it willy-nilly! No, the malware was delivered through a link while using webbrowsing. Gone are the days when you would be safe from viruses as long as you didn't manually run anything fishy yourself; nowadays, Flash and Javascript are almost mandatory when browsing the web. I don't think I will be deleting my system restore points, especially since it looks like it helped, not hurt, as I explained above! TresÁrboles (talk) 01:53, 15 January 2010 (UTC)
- On another note, I'd like to add that since your system restore points may now be infected, you should delete them.--Drknkn (talk) 13:16, 14 January 2010 (UTC)
- Looks like you havnt read what I wrote. Or perhaps you do not comprehend that you have to command the scanners to do a scan, and that when they are not doing a scan they are off. 78.147.87.28 (talk) 19:42, 14 January 2010 (UTC)
- No. Spybot, Search & Destroy comes with the "Tea Timer" which supposedly protects your system from unwanted changes. It has been known to interfere with other anti-malware programs. Make sure you didn't install that when you installed Spybot. Windows Defender also monitors your system for changes. The same goes for Avast! I've never used Malware Bytes or Super-Anti Spyware, though. I have Spybot S&D, too -- mostly because I'm used to it and have been using it for ages. It's not terribly sophisticated, but I made sure not to install Tea Timer when the installer asked me. I also use Kaspersky. I chose Kaspersky because it's easy to turn off. I rarely use either program, in fact. They mostly just sit there. I can't remember the last time I got an infection, but I work on computers for a living, so I have to disinfect other computers often. I just check for suspicious activity in the Task Manager and look for new files that run at startup. It's pretty obvious when you get an infection. Your computer usually slows down and programs start acting up out of the blue. This is a vast oversimplification, of course, but I'd nevertheless encourage you to research how to manually diagnose and disinfect your computer. I've saved my customers quite a bit of time and money by simply deleting the files and registry keys in question by hand rather than waiting for some slow program to do it.--Drknkn (talk) 20:50, 14 January 2010 (UTC)
- Well, dahhh! Don't use tea-timer or Windows Defender. If you want to avoid using more than one on-all-the-time program, then do not start up more than one on-all-the-time program. I thought that would be very obvious? You would have to positively choose to start them up. 84.13.50.207 (talk) 18:51, 15 January 2010 (UTC)
- No. Spybot, Search & Destroy comes with the "Tea Timer" which supposedly protects your system from unwanted changes. It has been known to interfere with other anti-malware programs. Make sure you didn't install that when you installed Spybot. Windows Defender also monitors your system for changes. The same goes for Avast! I've never used Malware Bytes or Super-Anti Spyware, though. I have Spybot S&D, too -- mostly because I'm used to it and have been using it for ages. It's not terribly sophisticated, but I made sure not to install Tea Timer when the installer asked me. I also use Kaspersky. I chose Kaspersky because it's easy to turn off. I rarely use either program, in fact. They mostly just sit there. I can't remember the last time I got an infection, but I work on computers for a living, so I have to disinfect other computers often. I just check for suspicious activity in the Task Manager and look for new files that run at startup. It's pretty obvious when you get an infection. Your computer usually slows down and programs start acting up out of the blue. This is a vast oversimplification, of course, but I'd nevertheless encourage you to research how to manually diagnose and disinfect your computer. I've saved my customers quite a bit of time and money by simply deleting the files and registry keys in question by hand rather than waiting for some slow program to do it.--Drknkn (talk) 20:50, 14 January 2010 (UTC)
- Looks like you havnt read what I wrote. Or perhaps you do not comprehend that you have to command the scanners to do a scan, and that when they are not doing a scan they are off. 78.147.87.28 (talk) 19:42, 14 January 2010 (UTC)
- Actually I'd go a step further and state that you should not trust antivirus software to save you from all .exe, .com, .bat, .vbs, and .js programs. Antivirus software is useless against, for example, a Trojan horse that some guy just wrote yesterday and it hasn't been dealt with yet by the antivirus software vendors. Don't download and execute any of those unless it's from a trusted source. If you wanted to go even a step further, you could use virtualization software like VirtualBox and do all your "risky" computing activity in the VirtualBox. If it turns out you had downloaded a Trojan, just delete the whole box and create a new one, and you're instantly clean. (Of course, until you determine that you had downloaded a Trojan, the VirtualBox may remain compromised, converted into a zombie computer, hammering your internal network, sending out zombie e-mails, etc., until you realize what has been happening.) Comet Tuttle (talk) 18:09, 14 January 2010 (UTC)
- I upload any suspicious files to http://www.virustotal.com. That site will scan it with 50 different anti-virus scanners. Heuristics often catch new viruses. .bat, .vbs, and .js files are plain text, so I just look at them in notepad.--Drknkn (talk) 20:50, 14 January 2010 (UTC)
Free RSS to email forwarding service
[edit]The RSS to email forwarding service that I used to use has sudenly decided to become a paid service. Does anyone know of an easy to use RSS forwarding service. I've done some searching but I'm unfortunatly not very tech savy and I havn't come up with much yet.
Thanks 196.213.43.74 (talk) 06:40, 12 January 2010 (UTC)
- gmail has web clips, but I don't if you can forward or pull them out of gmail.--194.197.235.240 (talk) 10:11, 14 January 2010 (UTC)
1GB vs GDDR5
[edit]I plan to buy a single-slot (single width) NVidia GeForce GT 240 video card. I have a choice of: GDDR3 with 1GB of memory, or GDDR5 with 512MB of memory. (GDDR5 with 1GB does not seem to be available in a single slot version.) Which should I get?
I don't play many (if any) 3D games. It's mainly for desktop use, but I want to try some of the new 3D window managers (Compiz), and I want to try CUDA. Ariel. (talk) 11:11, 14 January 2010 (UTC)
- For your needs, I do not think that you need more than 512 MB, so the faster GDDR5 memory may be the better choise. --Andreas Rejbrand (talk) 11:17, 14 January 2010 (UTC)
- If you're programming with CUDA, do some analysis and estimate the maximum problem size you will be able to compute for a given size of RAM. I have a GTS 250 for CUDA development and a cluster of T10 C1060 for real work; depending on your algorithm, you can burn through 512 MB with very small 3D problems. So, which will be a bigger problem for you - latency to GPU memory or total size of GPU memory? This is a problem-specific answer. For my work (solving the wave equation and reverse time migrationred-linked? Geophysical migration is close...), total size of GPU memory is more of a limiting factor than GPU global memory latency/throughput, and I usually stall on data transfers between the CPU and GPU (which is the fault of the PCIe bus, not the RAM type). Nimur (talk) 14:56, 14 January 2010 (UTC)
- (I have to admit that I did not see part of CUDA in the question. --Andreas Rejbrand (talk) 16:10, 14 January 2010 (UTC))
- If you're programming with CUDA, do some analysis and estimate the maximum problem size you will be able to compute for a given size of RAM. I have a GTS 250 for CUDA development and a cluster of T10 C1060 for real work; depending on your algorithm, you can burn through 512 MB with very small 3D problems. So, which will be a bigger problem for you - latency to GPU memory or total size of GPU memory? This is a problem-specific answer. For my work (solving the wave equation and reverse time migrationred-linked? Geophysical migration is close...), total size of GPU memory is more of a limiting factor than GPU global memory latency/throughput, and I usually stall on data transfers between the CPU and GPU (which is the fault of the PCIe bus, not the RAM type). Nimur (talk) 14:56, 14 January 2010 (UTC)
- Thanks! I was just going to play with CUDA, and learn about it. I have no specific problems I need to solve. Is there anything else more memory will help me with? I'm assuming 512MB is enough to do lots of things with CUDA? I was interested in trying out that CUDA cross compiler I heard of and running a program on the X86 and Nvidia at the same time. How much will the GDDR5 help with day to day 2D desktop usage? Ariel. (talk) 01:13, 15 January 2010 (UTC)
- Sure, you can do quite a lot with 512 MB. Let me reiterate for emphasis, that if you are doing true volumetric 3D problems, you will use up memory much faster than you think (... roughly speaking, per the cube of the problem size); but you can get by with a 50x50x50 grid for a lot of fun experiments. As far as 2D and 3D desktop effects - by the time you have a G80 or T10 processor (which you must, to be CUDA compliant), you are so dramatically far beyond the performance requirements that you really shouldn't have to worry about most gaming/2D desktop related stuff. One thing I have noticed is that time-sharing the GPU (e.g. running a 3D desktop manager like Compiz or Beryl) while computing with CUDA is potentially a bad idea. Sometimes this has kernel-panicked (crashed) my Debian or Ubuntu system. Sometimes it has made running X, (3D desktop manager or not), totally impossible on my RedHat systems. Upgrading to the newest driver has sequentially improved stability since Ubuntu 9.04. I have had luck with the last driver (cudadriver_2.3_linux_64_190.18, on Ubuntu with 2.6.31-16-generic kernel); no system instability. If you have a 64bit system, you may have some unique issues compiling the NVIDIA kernel modules (or they may work flawlessly). Finally, watch out for the latest Ubuntus, which come default with GCC 4.4, which is not CUDA-2.3 compliant. You will need to install a GCC-4.3 or older compatible version. Officially, the Ubuntu 9.04 is the last supported version, and one of the Enterprise linuxes such as Red Hat Enterprise Linux is more "officially" preferred. Nimur (talk) 01:58, 15 January 2010 (UTC)
- Sorry did you mean to say 'desktop/2D gaming' or '2D/3D desktop' or whatever instead of 'gaming/2D desktop' related stuff? Because a G240 while IMHO not bad, is definitely not something you would say you don't have to worry about it (I agree with the desktop POV however) if you are interested in modern 3D gaming which I appreciate the OP isn't (perhaps it's true on Linux since there are unfortunately fairly few modern 3D games for Linux so if you are gaming you're fairly limited in what's available) Nil Einne (talk) 11:58, 16 January 2010 (UTC)
- I don't know - are 3D games these days requiring more than 1 GB of video RAM and a 2 GHz GPU with more than 128 streaming processors? What exactly are they rendering? I admit that the newest addition to my games collection is at least 5 or 6 years old and I'm unfortunately not up to speed on the very latest recommended specs. But the cards and systems that are specifically marketed for CUDA are far and away the highest performing NVIDIA hardware (4 GPUs and 16 GB of video RAM on my S1070); and any of the gamer cards that are CUDA compliant are at least in the top series. Nimur (talk) 14:55, 16 January 2010 (UTC)
- 'Require' is a strong word. Most (if not all) games have to be able to run on less than that because if they did not, they'd sell too few copies to make money. However, there are plenty of games (notably, the ones based on the latest Crysis engine) that can make productive use of more processing power than any currently available graphics card can deliver. So sure, you can run almost all games with less horsepower than that - but there are plenty that will look simply gorgeous if you do have it. So games don't "require" that much - but that doesn't mean that you don't want to buy that much. Also, the fancier your graphics card - the longer it'll be before it's so obsolete that you won't be able to find any games that can run on it! SteveBaker (talk) 02:35, 17 January 2010 (UTC)
- Note I didn't say require and indeed I don't know of any game that requires a card better then GT 240. However many games will definitely benefit from it as SB says. I personally believe that people are too nuts about graphics and the need for a good graphics card, particularly in the developed world but I still prefer adventure games in general, and think that any game that requires excellent graphics to be enjoyable is by definition not a good game so I'm not exactly a typical gamer (and I do use a 8600GT). Also I have a CRT monitor, so am perfectly happy with low resolutions like 1024x768 whereas given the proliferation of LCD monitors, most people want their native resolution which will often be 1600x1080 or probably even higher like 1920x1200.
- Anyway I have no doubt that people using GPGPU can have some pretty massive requirements but this doesn't of course mean that a GT 240 is considered good enough that you don't have to worry about the card if you're in to gaming particularly if you don't intended to upgrade in a year. I would say a GT 240 is at best a mid range card when it comes to modern normal 3D gaming. For the more casual stuff like Sims 3, Spore, and particularly the indie or budget/small development team games of course the GT 240 would be plenty.
- The requirements in the gaming world are somewhat different from the GPGPU world. Generally speaking for a question like the OPs relating to the GPU RAM the answer would definitely be get the faster RAM rather then more RAM because you're much more likely to get a benefit from the faster RAM. Also while quad-SLI or quad-crossfire is popular among a small minority of gamers (and many gamers would probably like, if you were giving to it to them for free), I personally and many other gamers would prefer a better/faster single GPU card rather then either a multiple GPU card or multiple GPU set up (even one which is theoretically slightly faster) because while SLI and crossfire are resonably well supported particularly in the high end games, you're still far from guaranteed to get close to double the performance with two GPU. Note that in a multiple GPU setup for gaming, all content is duplicated as well, so if you have 4 cards with a total of 16GB ram, you only have the equivalent of 4GB of GPU RAM (which is still an excessive amount for gaming).
- Finally while I don't really pay that much attention to the high end GPU world, I believe the fastest current cards (perhaps until Nvidia launched the GeForce 300 series, their fastest card is the GTX 295 and it's probably fair to say they're fairly far behind at the moment) are the Radeon 5970s [1]. And while Crysis (and perhaps other games using the engine) is perhaps the only game where you're like to actually see much of a current benefit from that compared to some 'slower' cards [2] [3] particularly if you aren't using 2560x1600, we're talking about a card that's a very far cry from the GT 240, (note that this concurs with what I had expected, for gaming the DDR3 is a bad idea) [4] [5] [6] gives you an idea of where that sits. Note that for computer gaming, many people consider a ~30 FPS (average) around the minimum you want for decent gaming, ideally higher since this is an average.
- Anyway since the OP isn't interested in gaming, this is mostly OT so I made it small but since the issue came up, I felt it important to point out that a GT 240 definitely would often not be considered ' so dramatically far beyond the performance requirements that you really shouldn't have to worry'
- P.S. Note that some cards/vendors may specify GDDR3 when they really mean DDR3. According to [7] GDDR3 is supported but it wouldn't surprise me if many may choose to use DDR3 or GDDR5 not GDDR3.
- Nil Einne (talk) 07:25, 17 January 2010 (UTC)
- I don't know - are 3D games these days requiring more than 1 GB of video RAM and a 2 GHz GPU with more than 128 streaming processors? What exactly are they rendering? I admit that the newest addition to my games collection is at least 5 or 6 years old and I'm unfortunately not up to speed on the very latest recommended specs. But the cards and systems that are specifically marketed for CUDA are far and away the highest performing NVIDIA hardware (4 GPUs and 16 GB of video RAM on my S1070); and any of the gamer cards that are CUDA compliant are at least in the top series. Nimur (talk) 14:55, 16 January 2010 (UTC)
- Sorry did you mean to say 'desktop/2D gaming' or '2D/3D desktop' or whatever instead of 'gaming/2D desktop' related stuff? Because a G240 while IMHO not bad, is definitely not something you would say you don't have to worry about it (I agree with the desktop POV however) if you are interested in modern 3D gaming which I appreciate the OP isn't (perhaps it's true on Linux since there are unfortunately fairly few modern 3D games for Linux so if you are gaming you're fairly limited in what's available) Nil Einne (talk) 11:58, 16 January 2010 (UTC)
- Sure, you can do quite a lot with 512 MB. Let me reiterate for emphasis, that if you are doing true volumetric 3D problems, you will use up memory much faster than you think (... roughly speaking, per the cube of the problem size); but you can get by with a 50x50x50 grid for a lot of fun experiments. As far as 2D and 3D desktop effects - by the time you have a G80 or T10 processor (which you must, to be CUDA compliant), you are so dramatically far beyond the performance requirements that you really shouldn't have to worry about most gaming/2D desktop related stuff. One thing I have noticed is that time-sharing the GPU (e.g. running a 3D desktop manager like Compiz or Beryl) while computing with CUDA is potentially a bad idea. Sometimes this has kernel-panicked (crashed) my Debian or Ubuntu system. Sometimes it has made running X, (3D desktop manager or not), totally impossible on my RedHat systems. Upgrading to the newest driver has sequentially improved stability since Ubuntu 9.04. I have had luck with the last driver (cudadriver_2.3_linux_64_190.18, on Ubuntu with 2.6.31-16-generic kernel); no system instability. If you have a 64bit system, you may have some unique issues compiling the NVIDIA kernel modules (or they may work flawlessly). Finally, watch out for the latest Ubuntus, which come default with GCC 4.4, which is not CUDA-2.3 compliant. You will need to install a GCC-4.3 or older compatible version. Officially, the Ubuntu 9.04 is the last supported version, and one of the Enterprise linuxes such as Red Hat Enterprise Linux is more "officially" preferred. Nimur (talk) 01:58, 15 January 2010 (UTC)
Mail merge into objects and advance records in Word - VBA code?
[edit]I am trying to mail merge fields into several text boxes, but I want to advance the record in each one - however, I cannot put the <<NextRecord>> command in a text box, because Word does not allow it. Is there some other way around this? Or is there some VBA code I could use? Any help is appreciated! —Preceding unsigned comment added by 92.24.155.28 (talk) 14:08, 14 January 2010 (UTC)
old php
[edit]Will a php script written in php4 or even older php versions still work on a modern server? —Preceding unsigned comment added by 82.43.88.124 (talk) 18:32, 14 January 2010 (UTC)
- It depends on many factors. PHP5 is mostly back-compatible, but there was a major change to the Zend engine. Mostly, this affected behind-the-scenes performance, but that may have subtle effects on "weird" things. I specifically remember trouble with the GD library which is commonly used for graphics and images (e.g. if you have a thumbnail generator script, you may have to tweak or rewrite it). Take a look at the PHP 5.0.0 changelog and this important document, Migrating from PHP 4 to PHP 5. Watch for subtle changes in any of your custom configuration scripts if they refer to specific engine/PHP language levels. Nimur (talk) 20:28, 14 January 2010 (UTC)
Function keys
[edit]I often hear Function keys referred to as "PFX" where X equals the number. I know that F means function but what did the P stand for? I've been told the "P" is a bit of IBM jargon but I haven't been able to find anyone who knows that the P ever stood for. Thanks, Dismas|(talk) 20:06, 14 January 2010 (UTC)
- IBM 3270 calls them "Programmed Function Keys". -- Coneslayer (talk) 20:12, 14 January 2010 (UTC)
- Back on my IBM Personal System/2 (the real PS2), I recall that these were actually labeled with a "PF" on the keyboard. They were also on the left hand side of the keyboard, not across the top. Many of them had software-specific labels as well, e.g. "copy" "paste" and "spell-check" which were designed to directly map on to Microsoft Edit or some other console-mode word processor. (I believe it may even have been called Word, but it's been some 20 years since I used that program...) The closest image I can find on the web is this Japanese IBM PS/2 keyboard, which has both F-keys and PF-keys; but mine was a US version. Nimur (talk) 20:31, 14 January 2010 (UTC)
- Correction - I think it was pfs:Write, not Word. This doesn't appear to have ever been a Microsoft product... Nimur (talk) 20:39, 14 January 2010 (UTC)
- Back on my IBM Personal System/2 (the real PS2), I recall that these were actually labeled with a "PF" on the keyboard. They were also on the left hand side of the keyboard, not across the top. Many of them had software-specific labels as well, e.g. "copy" "paste" and "spell-check" which were designed to directly map on to Microsoft Edit or some other console-mode word processor. (I believe it may even have been called Word, but it's been some 20 years since I used that program...) The closest image I can find on the web is this Japanese IBM PS/2 keyboard, which has both F-keys and PF-keys; but mine was a US version. Nimur (talk) 20:31, 14 January 2010 (UTC)
- Thanks all! Dismas|(talk) 07:39, 17 January 2010 (UTC)
neural network integrated circuits
[edit]I've found lots of research lab references to neural network integrated circuits but no references to actual chips out on the market for sale. Where can I find neural network chips that are available for purchase? --71.100.14.125 (talk) 22:17, 14 January 2010 (UTC)
- Is this one of any interest? I have no particular knowledge here; I just googled "neural network" chip and then clicked "Shopping" at the top of the resulting Google page. Comet Tuttle (talk) 22:41, 14 January 2010 (UTC)
- You can implement a neural network on an FPGA. Those are widely available and can be much cheaper than the special-purpose chip linked above. Nimur (talk) 04:49, 15 January 2010 (UTC)
Photoshop and PNGs
[edit]My question is about Adobe Photoshop and its handling of alpha channels of PNG files. If I load an RGB graphics file that has an alpha channel, like a TGA or PSD file, then Photoshop's "Channels" tab shows the channels "RGB", "Red", "Green", "Blue", and "Alpha 1", as expected. When I save this file as a PNG and then load the PNG into Photoshop, the "Alpha 1" channel doesn't exist anymore under the "Channels" tab. The transparent areas of the bitmap are still shown as transparent, but there is no alpha channel in the Channels tab for me to edit. How do I get to the alpha channel in Photoshop? (Pre-emptive notes: This behavior has been observed both in Photoshop CS2 and CS4. I need to do this in Photoshop, not the GIMP or any other graphics utility. I've tried using the SuperPNG plugin instead of the PNG plugin that comes with Photoshop, and there's no change in the behavior of Photoshop.) Comet Tuttle (talk) 22:40, 14 January 2010 (UTC)
- Followup question in case I was not clear enough above: Is there a way to make it do what I want (store alpha channel information in a separate channel as listed under the "Channels" tab)? Comet Tuttle (talk) 22:56, 14 January 2010 (UTC)
- Reading our Portable Network Graphics article, I am just coming to learn that PNG files actually have numerous options for storing transparency information. Unfortunately, while alpha channels are a valid option according to the PNG specification, they don't seem to be the standard and are not commonly supported in a lot of graphics tools. Instead, a palette-ized "special pixel code", representing transparency, is more common. It would seem that the default library that Photoshop uses does not support saving PNG files with complete Alpha Channel transparency, opting instead for one of these palette-ized variants. I haven't found a workaround. Nimur (talk) 02:28, 15 January 2010 (UTC)
- That's so weird! I'm deeply confused by Nimur's response. I work in computer graphics - we use transparency in PNG all the time - it's our primary archive format for texture maps - I've been doing that for well over 10 years and I've never once had problems in either Photoshop or GIMP!! If Nimur wasn't someone I trust, I'd call "bullshit"...but I'm not going to do that. I think that all of this weirdness that you refer to must result from using palletized PNG. When you palletize an image, you lose quality - and the quality loss in alpha is (evidently) dramatic. We never use the palleted version of PNG - so I'm pretty sure that if you save them as full color then there is only one way that alpha is stored and it always works just fine - even with the stock Photoshop loader. Older versions of Internet Explorer have trouble displaying transparent PNG - perhaps that's what's confusing you?
- SteveBaker (talk) 03:34, 15 January 2010 (UTC)
- Let me clarify my poorly worded explanation. First of all, for the record, I love PNG. They are among the most versatile tool for web-based imagery, and suitable for photo-quality or diagram-style images, they preserve text - etc. They have numerous benefits. Most importantly, their transparency support is spectacular (discounting the old browser versions Steve mentioned). But what I am surprised to discover is that most PNG transparency is not performed by storing a separate alpha channel. It seems that in my version of Photoshop (Photoshop 7, albeit a little old), it is not even an option to store a complete alpha channel for a PNG. So, if you are exporting a PNG file with transparent pixels, Photoshop is always palette-izing it. From Comet Tuttle's description, it seems that this is the case in even newer Photoshop CS 2 and 4.
- This is very unfortunate, because you lose the complete 8-bit/256-level transparency capability; and you lose some of your total dynamic range in the rest of the color channels. This is further complicated by the compression that PNG applies. If Steve or anyone else does know of a way to export PNG from Photoshop with full alpha channel (e.g. a true 32-bit per pixel, RGBa scheme)), by all means please post a link or a description. Nimur (talk) 03:47, 15 January 2010 (UTC)
- Addendum - since this is the reference desk, let me link to the official PNG Specification: tRNS Transparency vs. Alpha Channel transparency. Clearly the superior quality of a separate alpha channel is counterbalanced against the superior compression capability of tRNS. The introduction to transparency discusses technical details. There is also support for both pre- and non-premultiplied alpha. I don't know why a professional image editing suite like Photoshop would not provide an option to control this detail - it's something a professional image editor would care about. Nimur (talk) 04:02, 15 January 2010 (UTC)
- Further experimentation is showing even more complexity. My best guess is that Photoshop zeros out any pixel with 0 alpha, before exporting to PNG and compressing. I've been experimenting with an alpha channel on Liz Phair, starting with an original, creating an alpha channel, ranging from 0.0 to 1.0, in Photoshop and exporting as a PNG with transparency, and then after re-loading that result into Photoshop or GIMP, I attempt to reconstruct the original by re-setting the alpha to 1.0 every pixel - but although some pixels are restored, it doesn't reconstruct everything. Clearly, there's a mess of lossy compression, block coding, alpha pre-multiplication, and inter-channel compression, etc., that is beyond the simple approximation of separate channels for each color and alpha. Short of a bitmap format or totally lossless compression, I think this is as good as it gets. Nimur (talk) 04:36, 15 January 2010 (UTC)
- There is a Photoshop plugin from 2007 called SuperPNG from 2007, and another plugin called PNGOUT, both of which purport to improve PNG compatibility from the built-in Adobe solution, and the latter specifically states that it can be configured to write PNGs with 24-bit color with an 8-bit alpha channel. But this is slightly beside the point — what has baffled me is why Photoshop doesn't load any variety of PNG, no matter how it stores transparency, and stores the transparency information as an alpha channel, to support the standard workflow that Photoshop artists are accustomed to. I have not found a PNG file yet that Photoshop (for Windows, CS2 and CS4, anyway) loads and displays an alpha channel for. Comet Tuttle (talk) 05:56, 15 January 2010 (UTC)
- Further experimentation is showing even more complexity. My best guess is that Photoshop zeros out any pixel with 0 alpha, before exporting to PNG and compressing. I've been experimenting with an alpha channel on Liz Phair, starting with an original, creating an alpha channel, ranging from 0.0 to 1.0, in Photoshop and exporting as a PNG with transparency, and then after re-loading that result into Photoshop or GIMP, I attempt to reconstruct the original by re-setting the alpha to 1.0 every pixel - but although some pixels are restored, it doesn't reconstruct everything. Clearly, there's a mess of lossy compression, block coding, alpha pre-multiplication, and inter-channel compression, etc., that is beyond the simple approximation of separate channels for each color and alpha. Short of a bitmap format or totally lossless compression, I think this is as good as it gets. Nimur (talk) 04:36, 15 January 2010 (UTC)
- Addendum - since this is the reference desk, let me link to the official PNG Specification: tRNS Transparency vs. Alpha Channel transparency. Clearly the superior quality of a separate alpha channel is counterbalanced against the superior compression capability of tRNS. The introduction to transparency discusses technical details. There is also support for both pre- and non-premultiplied alpha. I don't know why a professional image editing suite like Photoshop would not provide an option to control this detail - it's something a professional image editor would care about. Nimur (talk) 04:02, 15 January 2010 (UTC)
LaTeX textstyle in a region
[edit]In LaTeX, is there any way to apply \textstyle
to some region (say an entire \align
), or to the entire document? 128.237.251.92 (talk) 22:43, 14 January 2010 (UTC)
- You can use
\DeclareMathSizes
, but that applies only to the entire document and changes only the size of symbols, not other things like limits placement. I don't know of a way to extend\textstyle
's reach, sorry. --Tardis (talk) 16:28, 15 January 2010 (UTC)- Define a new command that combines the \align with the \textstyle. Taemyr (talk) 08:51, 19 January 2010 (UTC)
- Unfortunately, a
\begin{align}
block may need multiple\textstyle
declarations within it, because their scope is limited (there's no "toplevel" inside the align where you could put it). You'd have to do some really clever redefinition so that things like&
within your command/environment generated their own\textstyle
s. By contrast (and as an extension to my original answer), you can put an entire\begin{aligned}
(fromamsmath
) under the domain of a\textstyle
, but that's somewhat different. --Tardis (talk) 16:49, 19 January 2010 (UTC)
- Unfortunately, a
- Define a new command that combines the \align with the \textstyle. Taemyr (talk) 08:51, 19 January 2010 (UTC)