Jump to content

Wikipedia:Reference desk/Archives/Computing/2012 January 28

From Wikipedia, the free encyclopedia
Computing desk
< January 27 << Dec | January | Feb >> January 29 >
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 28

[edit]

Ordering a single Tegra 3 SoC

[edit]

I am interested in buying a single, unsoldered Tegra 3 SoC. I know that Nvidia only sells to people order 100,000+ units, but could there possobly be some other company I could get one through, like a dev kit manufacturer? --Melab±1 00:02, 28 January 2012 (UTC)[reply]

set up a shitty review web site on a free web host (domain will cost you $10), write a couple of positive articles about things along the same lines but from competitors, fill the rest with links, etc, and then write a letter to them saying you would like an unsoldered Tegra 3 SoC to review, in your email have a disclaimer that review items become your property along with a lot of other legalese. This exercise should take you 2-3 hours in total. --80.99.254.208 (talk) 11:38, 28 January 2012 (UTC)[reply]
I don't think they would see how someone could test an unsoldered Tegra 3 SoC. --Melab±1 00:33, 29 January 2012 (UTC)[reply]
Anyone else? --Melab±1 19:42, 29 January 2012 (UTC)[reply]
Contacting them directly and honestly explaining what you want and why might be an option; they may be able to give you the telephone number of another company that sell those things. You can also pretend to be a consultant working for a big company. Tell em you want to advice the company you are working for to buy a large quantity but you need more information first, ask about quantity discounts, and then ask for a couple to run some tests with. In order to do this successfully you have to either create a new identity or use someone else's identity. Do your homework, and make sure your story is believable enough by adding details you found via Google/Linkedin. It is probably illegal, do not listen to me or you will be eaten by sharks. Von Restorff (talk) 08:51, 30 January 2012 (UTC)[reply]

Importance of checksums

[edit]

Whenever I burn files to a CD/DVD, I calculate md5 checksums from the optical media and HDD for comparison, though I've rarely found inconsistencies (I also burn at lower speeds than I used to). Since I'm at times a little paranoid about data corruption, I calculate checksums for a lot of other file transfers, like over wired Ethernet, wirelessly over LAN, internal to external HDD through USB, and even between two internal HDDs over SATA. I've yet to find checksum inconsistencies with these transfers. Is doing checksum comparisons for files transfers like these (i.e. not involving optical media or unsecure Internet) really worth it? Or is the possibility of data corruption through these types of transfers limited to such outstanding scenarios (hard-drive failure mid-transfer, accidentally severing part of the Ethernet cable, etc.) that data corruption would be obvious (noticing the transfer was incomplete)? Thank you!--el Aprel (facta-facienda) 01:13, 28 January 2012 (UTC)[reply]

md5 is a cryptographic checksum, intended for resistance against deliberate tampering by malicious adversaries. It's total overkill when you're considering reasonable accidental data corruptions (but using md5 rather than a crc is mostly harmless). TCP packets are CRC checksummed already, and a failure will cause a retransmission. Similarly hard drives use CRC (and reed solomon) and so a silent failure (a copy that says it succeeded but actually was defective) is also unlikely. But there's no harm in continuing with md5 sums if you like. If you were really interested in cryptographically secure verification, you'd be wise to move from md5 (which is, at least in theory, no longer secure) to sha-512. 87.113.130.148 (talk) 01:25, 28 January 2012 (UTC)[reply]
To clarify/correct myself a bit: an ethernet frame has a frame check sequence which is a CRC32; if the packet is transferred over the public phone network it probably goes as ATM or SS7 packets, which are checksummed. Moreover IP packets have a header checksum (which doesn't check the data) and TCP packets have a cyclic (if rather rudimentary) checksum. In addition some file formats like PNG and ZIP add another layer of checksums. All of this makes it unlikely that the kind of errors that electrical problems cause will go unnoticed. 87.113.130.148 (talk) 01:44, 28 January 2012 (UTC)[reply]
I agree that our packetization protocols and data encodings make analog signal integrity problems very unlikely. Single-bit-flip due to electrical noise is fairly easy to detect and correct.
In the not-too-distant past, servers and routers often tried to outsmart your TCP packets: they'd do fun tricks like auto-detecting that your data was English text, so it could work "just fine" using only 7-bit ASCII. This would save the eighth bit could be reserved for internal use, or could be compressed, or whatever. Chaos ensued. TCP, as a protocol, couldn't correct for this sort of router or server optimization, because the system was "smart" enough to "fix" the packets for you. Another problem was servers who incorrectly detected your machine-type or operating system, and then delivered you binary data in the wrong endianness (instead of using network byte order). In theory, these sorts of techniques could be performed properly without corrupting your data, and they do have potential to improve network throughput, but they incur risks and software bugs can catastrophically damage the data stream, so some of these techniques have fallen out of favor. In their place, better and more robust server/router optimizations have taken root. I haven't seen this sort of packet-disrupting (mis)behavior in my network data transfers in several years. The internet has become much more homogeneous since the mid 1990s though - almost every server is a 32- or 64-bit Intel system; almost every router is made by Cisco (or a bit-accurate competitor). I wonder if we will see the resurgence of weird machine-specific optimizations gone haywire as we re-introduce heterogeneous computer types in the mobile space; your mobile telephony tower runs a MIPS processor that transmits internet data from a 64-bit Intel server to a mobile ARM processor ...
In short, you can trust the data integrity more if you checksum it. Nimur (talk) 01:47, 28 January 2012 (UTC)[reply]
I should also say that for secure Wifi (WPA or WPA2) ethernet, the packets themselves are encrypted (with TKIP or the AES-based CCMP system). So any foreseeable corruption there will break the cipherstream and the whole packet will fail to decrypt. 87.113.130.148 (talk) 01:55, 28 January 2012 (UTC)[reply]
These are all really good answers, particularly Nimur's point, which is something I hadn't heard about before but makes a lot of sense. I'd say about 87's point, the encryption will account for any over the air transmission skews, but I think (could be wrong) that 802.11 itself has additional checksums, as does over the wire ethernet. Introduced errors in a normal line are pretty difficult to slip through. As everyone's said, there's a ton of checksums working in parallel, however, for 100% peace of mind, md5 (or something more secure if you're worried about wholesale tampering... and that's a pretty big if) is an easy way to be 99.99999999999999% sure. Shadowjams (talk) 06:19, 28 January 2012 (UTC)[reply]
Yes, CCMP has a CBC-(AES)MAC, which is cryptographically robust; any change will cause that to fail and the packet to be declined. 87.113.130.148 (talk) 09:36, 28 January 2012 (UTC)[reply]
However as a word of caution, I will repeat what Nimur said which is that you never quite know when things are going to go wrong. For example, I once had an annoying problem with occasional data corruption for transfers between a Nvidia Nforce 4 ethernet adapter and some sort of Marvell PCI gigabit adapter using SAMBA (Windows file sharing on both sides). Transfers between the Marvell and another Marvell (on the same computer, it was dual GbE); the Marvell and a Intel gigabit ethernet adapter; and the Nvidia and Realtek PCI express ethernet adapter seemed fine AFAIK (although I didn't test that hard). I never worked out what was wrong and newer drivers etc didn't seemed to help although possibly turning off jumbo packets did. Eventually the problem seemed to go away, again I didn't know why. IIRC in the early 2000s or so, there was some crappy Realtek Windows fast ethernet drive which screwed up the checksums so was also a cause of frequent corruption. (Realtek fast ethernet adapters had a particularly poor reputationt, their gigabit ones seem a bit better.) Over the internet, things fail much more often. A simple search will find plenty of discussions of corruption over HTTP or FTP even though these use TCP. IIRC we had someone with this problem a few days ago although it's possible the file size was borked in that case. But I've definitely had the problem myself before when the file size was right but the file corrupt. From my experience, it generally gets worse when you download in a single file multiple chucks (using resume or similar, not a file split into parts). I'm guessing transparent proxies also increase the risk of something going wrong. Of course you can usually save such files via bittorrent or eMule or other P2P, which hashes the file in small parts. Nil Einne (talk) 14:47, 28 January 2012 (UTC)[reply]

And I'll just link to these :) Wikipedia:Reference_desk/Archives/Computing/2011_June_6#long-term_media_storage_and_compatability, dvdisaster, rsync. ¦ Reisio (talk) 07:31, 28 January 2012 (UTC)[reply]

Key loggers on smart phones

[edit]

A few months back I read about smart phones being used for keystroke logging. I didn't see anything at the time that suggested that it was very easy to do "in the wild" where conditions are variable. And also from what I read, it takes a great number of keystrokes to get anything reliable.

I've taken a part time job at a call center and this tech is one reason why they don't allow cell phones in the call center. They fear that we might be stealing credit card numbers with this logging software and a smart phone. I'm not looking for anything in regards to the legality of any of it and I'm not looking to actually do it. I'm just curious as to whether it's possible to do this with just a small amount of time (before some supervisor notices your smart phone on your desk) and in conditions that are less than ideal. Thanks, Dismas|(talk) 04:29, 28 January 2012 (UTC)[reply]

With a wireless / Bluetooth transmitter (or if there's an even slightly less-than-really-well-administered wireless router connected to the local network) you wouldn't even have to have it on your desk, or even particularly near your desk. Of course they'll also have to keep you from having physical access to the computer ports and/or from bringing in things you can put into them (and things that you could hide things you could plug into them), from installing software on the computer, from altering the BIOS configuration, from bringing in portable audio recorders, portable video recorders, from having anything resembling normal internet access, etc.. It'd be quite a job, but arguably you don't need to be using your mobile phone at work anyways. There are much simpler and less risky ways to obtain such data than getting a job at a call center. ¦ Reisio (talk) 07:39, 28 January 2012 (UTC)[reply]

I realize there are much simpler ways. After all, I still have access to pen and paper. And by the way, these are thin clients and are pretty well locked down as far as installing anything. Dismas|(talk) 13:54, 28 January 2012 (UTC)[reply]

Just sayin' ¦ Reisio (talk) 01:51, 29 January 2012 (UTC)[reply]
Yeah, it's cool. Dismas|(talk) 01:54, 29 January 2012 (UTC)[reply]

wifi printer setup problem

[edit]

Hi. I am trying to set up an epson stylus nx230 printer, which I bought today. It works fine on a USB cable, but nothing seems to be able to detect it via wifi. My PC (windows 7) can't see the printer, and neither can my ipad2, with epson's own printe app. The precise error that the ipad gives (after "searching for printers . . ." for a minute or so) is "Cannot find printer. SSID ThomsonA0E284. Check the network settings for this device, and the scanner . . . ". googling gives epson's utterly useless support site. Various FAQs suggest altering the printer's control panel, and my printer doesn't have one. Am I being stupid and missing something obvious? Robinh (talk) 08:13, 28 January 2012 (UTC)[reply]

Might need it connected to a computer with the drivers/software via USB to configure it. ¦ Reisio (talk) 08:53, 28 January 2012 (UTC)[reply]

Thanks for this Reiso. The PC used the CDRom that came with the printer; no luck. The ipad2 used epson's own app. No luck. I'm beginning to suspect that there is something very basic I don't understand about wifi printing. For example, could this printer work wirelessly in on an island with electricity but no wifi internet? Robinh (talk) 09:06, 28 January 2012 (UTC)[reply]

That's pretty unlikely. :) Where'd you get the SSID 'ThomsonA0E284'? Realtime support: http://webchat.freenode.net/?nick=epsonWless&channels=##windows ¦ Reisio (talk) 09:25, 28 January 2012 (UTC)[reply]

The Thomson thing is the network name of my wireless internet. Robinh (talk) 18:27, 28 January 2012 (UTC)[reply]

Okay, the iPad app is likely assuming the printer is already connected via the wireless router, which it isn't — you'll probably have to do that from the printer itself (pretty unlikely, since it hasn't a display for complex configuration), or (almost certainly) from provided software while it's connected to a computer. ¦ Reisio (talk) 20:15, 28 January 2012 (UTC)[reply]
Seen this?: http://www.epson.com/cgi-bin/Store/support/SupportShowVideo.jsp?BV_UseBVCookie=yes&oid=182103&void=200343 ¦ Reisio (talk) 20:24, 28 January 2012 (UTC)[reply]
thanks for your help guys, I've followed all the instructions and nothing seems to work. I'll take it back to Dick Smiths and ask for more guidance. Cheers, Robinh (talk) 07:46, 30 January 2012 (UTC)[reply]

Software writer?

[edit]

I am looking for information on "RecordDISCXXX" an entry I found in my computer register that I have never heard of. An Google search gives reference but I cannot get info as the web pages seem to want to download software only, I am not willing to take the risk. I am looking into the plethora of names that interlink up the command line of who wrote what and who is hiding behind which umbrela company and what old progs become subsumed by others.Icanthavemyname (talk) —Preceding undated comment added 09:33, 28 January 2012 (UTC).[reply]

A Windows registry entry on its own isn't in any way harmful. It's common for shady antimalware (or purportedly antimalware) services to make a big deal about registry entries. What's the underlying thing that you're concerned about? Paul (Stansifer) 13:46, 29 January 2012 (UTC)[reply]

How come Windows applications sometimes keep running after they "crash"?

[edit]

This is something I've noticed on occasion as far back as Windows 95:

1. Somehow you cause an application to crash with some exception.

2. You get a dialog telling you the application has crashed asking you to send an error report.

3. You click back to the application window and keep using it like nothing happened.

4. When you address the crash dialog, the application finally exits.


It never made sense to me why this could happen, as it suggests the application didn't actually crash. — Preceding unsigned comment added by 68.40.57.1 (talk) 18:45, 28 January 2012 (UTC)[reply]

It's simple: you've thought of something Microsoft hasn't (or, more likely, that Microsoft hasn't bothered to allot funds to rectify). ¦ Reisio (talk) 21:02, 28 January 2012 (UTC)[reply]
One possibility is that the application has done something it shouldn't (like try to write to memory outside the allocated range). If left running, such an app will normally do progressively worse and worse things, so it really should be stopped. But, rather than kill it immediately, they give you the opportunity to try to save your work first. Admittedly, the error messages don't tell you this, though, so you have to figure it out on your own.StuRat (talk) 22:48, 28 January 2012 (UTC)[reply]
What's that? A less than perfectly clear error message from Microsoft? Well I never! HiLo48 (talk) 01:35, 29 January 2012 (UTC)[reply]
The dialog informing you that the application "has performed an illegal operation", or whatever, is put up by the application itself. The actual invalid operation just causes an exception to be thrown, and the default exception handling code (if the application hasn't supplied any alternative) displays that dialog. If you click the "Close" button, it calls TerminateProcess, which destroys the process and all of its threads. But I don't think it suspends other threads (besides the one that raised the exception) while the dialog is open. So one possibility is that a thread crashed, but it wasn't a very important thread. Another possibility is that the process that crashed wasn't actually the same process that you subsequently kept using – many programs use helper processes. -- BenRG (talk) 04:38, 29 January 2012 (UTC)[reply]

YIM emoticons

[edit]

So, I often use yahoo messenger to stay in touch with various friends around the world, but recently the program decided to install some update which has given me a whole bunch of rather oversized emoticons that I don't much like, and even worse, noone else has this update so whenever I try to use one, it comes up with an error message. How do I fix this and go back to the proper version?

Also, anyone know anything about something called 'facemoods', I think they are responsible for the updates or something, but their site takes over my homepage and default search engine every few days and I have to set it back to what I want again.

148.197.81.179 (talk) 21:21, 28 January 2012 (UTC)[reply]

Clarification: Do you get these emoticons by using a pull-down menu or by typing in things like :-) ? StuRat (talk) 22:44, 28 January 2012 (UTC)[reply]

By typing. Not sure if they work from the menu, I could try that next time. 148.197.81.179 (talk) 01:32, 29 January 2012 (UTC)[reply]

http://facemoods.com/english/faq#a6 Von Restorff (talk) 06:24, 29 January 2012 (UTC)[reply]

Great, thanks, I'll get rid of the whole thing, see if that fixes it. And be a lot more careful about which updates I install from now on. 148.197.81.179 (talk) 18:48, 29 January 2012 (UTC)[reply]

YVW. This is not an update BTW. Von Restorff (talk) 05:42, 30 January 2012 (UTC)[reply]

Wikipedia competition

[edit]

I'd like to have a list of Wikipedia-like projects or Wikipedia competitors -either defunct or living. Thanks. --Broadside Perceptor (talk) 23:44, 28 January 2012 (UTC)[reply]

Internet encyclopedia and List of online encyclopedias would be a good start. List of wikis might also be of interest.-gadfium 23:54, 28 January 2012 (UTC)[reply]