You must not be playing anything all that great, then. Anything with 5.1 or even sometimes 2.1 surround typically fails, higher bitrate stuff is often choppy, and often.mkv stuff won't play properly or allow you to pause/resume/seek. This is with the built-in video stuff as well as players like RockPlayer or MoboPlayer.
Microsoft's intent has been for sometime to remove their dependence on other vendors. What makes you think anything more than a thin client will be needed in a couple years? That is quite obviously the direct MS is going with Windows 8. Windows 8 + an ARM thinclient + MS terminal service applications. All the better if they're Office 365 applications, right?
You're correct; however, if I can detect someone's 2.4GHz wireless security camera(s), it's another story. It really would be an invasion of privacy at that point, assuming I'm intent upon invading your privacy.
Had I known this 10 years ago when I started going to school to get my degree, I'd have gone into pre-law, and be able to retire in just a couple years after a successful career in software patent litigation...
Yeah, for infrastructure/interconnects/uplinks I'll use pink or red, and keep that out of my 'bundling', if it makes sense for the environment. But otherwise, it's a grab bag of different colors.
The labeling isn't there to make tracing the cable easy, it's there to keep track of which cable end is plugged into which port on which device.
No, it does both. The label assists with tracing, which labeling the 'correct' side doesn't do (telling you where that cable is plugged in on the other side). It's the only thing you can't tell by labeling the side 'correctly'.
Meanwhile, you can still look at a cable that's plugged into a device and see "oh, hey, this cable is plugged into the device". Labeling it with that device's name saves you no time, and a situation where a cable would be unplugged on one end would be fairly easy to detect from either side. (Labeling the opposite end tells you where you need to look to figure out which machine is unplugged, however.)
Labeling 'correctly' serves no better function, while delivering less.
During the later days of floppy drives, the quality of the drives themselves went down. I'm not sure what it was, but they were less able to read and write floppies than the older drives. I still have some 3.5" floppy drives sitting in my 'recovery' (along with a board/etc. that can use it) which served me well. TEAC brand, IIRC. Both came out of ancient Gateway 2000 machines (486 and/or early Pentium towers, probably).
Don't use the USB/floppy converters/drives. They're almost worthless. Once floppies got around to about $0.50/ea, the drives themselves were nearly as useless.
From my recollection, dd on linux would seek until it got a good reading. More often than not, most of what was recovered was usable even with that. Failing that, there's also ddrescue and the other tools being recommended. (Linux also seems more able to read floppies than Windows or Mac, for what it's worth. I was usually able to 'recover' people's data from the disk simply by copying it to hard drive and re-writing it to another floppy.)
I would argue that there's really no good reason why the later android builds can't be made to work on the older phones. There are a number of Android phones with some fairly minimal specifications (500MHz CPU, 128-256Mb of RAM) which are quite responsive with the 2.3.x android build, and there are a number of phones that perform poorly simply due to the specific software build, not necessarily the "newness" or "featureful" nature of the software.
As an HTC HD2 owner, having tried a number of different ROMs, I've noticed that there is a huge variance in performance even on the same basic kernel/android version due to how the ROM was set up - everything from the specific 'tweaks' on how dalvik/caching/CPU is configured to which managers and widgets are used.
* primarily uses Exchange, which is (in its online/prevailant incarnation of bpos/office365) incompatible with anything but MS clients * has a mobile MS Office built-in (some call that a feature, and I'll admit, OneNote is sweet.) * does not release their source code
Android: * can use a myriad of mail services, including gmail, from which I can export all my mail from any client, of which there are several. I can also export my contacts in any format I please (literally) * has a number of available word processor type things. * releases their source code (for the most recent phone release, which I think is 2.3.4 at this point) for everything not dealing directly with their infrastructure * allows 3rd party software to be installed w/o boostrapping to their market (hmm, you weren't thinking winmo6.5 were you?)
Unlike "before", when there were some fairly interesting and innovative ideas on the mobile realm from open source in both hardware and software, there's actually a fair amount of R&D competition. Nokia's N800, N880, N900 and such, on Maemo, were open source to their core, with some fairly solid design philosophies. That was about it in terms of "open source mobile", and it had a fair amount of effort behind it (it, as well as fltk, tinyX, etc. on which it was based, and then later, Qt).
Now, however, you can actually get a job to work on this shit. Once taht's done, however, you're likely to have an NDA/noncomp which would inhibit your ability to work on OSS... oh well. (They're likely making serious bank in a down economy; don't jinx that shit.)
No kidding. They should just be happy to humiliate women for the humiliating treatment of women. It would be greatly preferable to the joke that the refer to as "ethical" treatment of animals.
(I miss being able to afford fur coats. Anyone remember back when you could actually buy a mink, fox, or coyote jacket for the winter without importing it from Russia or China?)
Your last point is easily mitigated by taking your second to last point a step further:
Lastly, limit access to the wiring closets only to those that need it, have been trained, and are held responsible if it becomes a mess.
You can use different cable colors for identifying certain things in your environment (wireless, printers, servers, etc). If you can't justify buying all of the cables sizes you need in all of the different colors then you can use colored tape or some other type of identifier like plastic tags. You have plenty of options.
What I propose and do myself is: don't care what color the cables are, as long as they're different. Then label them. Or, don't label them, but having different color cables in each bundle makes the "figure out where cables go" problem so very much easier and almost completely avoidable. You still have to trace cables with colored cables, and when they're colored by role, you'll often have bundles of same-colored cables. That's no good. Just mix and match, who gives a damn (the different colors per bundle make it easier to find and identify bundles as unique, too). You don't need to 'train' anyone to your specific environment for little things, and it makes it much easier to hand off to someone else.
At first I thought you were being sarcastic with a statement like this:
Proper use of cable ties means you are not afraid to use a LOT of them, and also not afraid of cutting them open when you need to change someting. I keep a cheap diagonal cutter and a bag of assorted cable ties in every desk drawer in my house (3 "kits" in total).
That makes sense in some applications, so maybe they do for your's. You weren't really specific.
However, let me just say that as someone who has come after the dipshits who cinch cable tied an entire rack of equipment in (and the people who came after them, who tried to make due with having all their cables constrained and inaccessible), this is a bad practice. Not only does it result in marginal forethought, but it requires two things you do not want:
1) You need an extra tool (and more time) to redo wiring or even rack another machine, sometimes. 2) That you have a knife near networking cable which is able to cut through the ties - ties which are more difficult to cut than the networking cable itself.
Case in point for #2: we were reorganizing a rack (removing every third/other system, moving the remaining machines down, and putting new equipment in the top). We were almost done and checking to make sure we didn't jiggle something loose or fail to plug something back in properly, and we noticed that the main DC was intermittently available. Someone had managed to nick the cable in a bundle when separating it and didn't notice - slicing through the cable enough to cause some fairly severe (but not complete) packet loss.
The ONLY time I would say a network/server rack should be cinch tied is when you are building a handful of racks to a universal specification and will not be touching the cable until the cable or equipment is completely replaced.
Do not use this shit - either the cable wrap or the cinch ties - in a server room. Please.
It may look nice, and for applications where "this looks really neatly done", that's important. A server room or a networking rack is not one of these places. In such places, it will only get in the way, turning a quick job into a lengthy one.
Like any addiction, the first step is to admit that you've got a problem.:)
Seriously, though. You've got to realize that whatever the result of your efforts is, it's still going to be non-ideal, even if it looks as uniform and as neat as humanly possible. The truth is, no matter of initial neatness will prevent someone from coming along and fucking it up, whether unintentionally and with good cause or because they're a lazy git.
At home, I will label both ends of a cable that is likely to be sitting in a pile with other cables (usually - particularly when they're likely to be confused with others and when tracing the cable will be difficult). Just a simple file folder label or piece of tape, usually. Mostly, I don't worry about it, because at home "oops, I unplugged the system" only bothers me and/or my family. At work, it's another story: everything you do should be done to help mitigate downtime and improve your ability to effectively work with the cabling.
On server racks, I prefer a handful of techniques. There are a couple principles I abide by:
* always assume the rack will get messy over time * your original intent will not be the intent of the next person to come along * never assume a standard, because everyone has a different idea of how things should be done, and first impressions to that effect can be wrong * standards only make sense when the implementation requirements/specifications/etc. are identical/universal, eg. with electrical wiring or in a large hosting facility where everything is the same or there is a standard which can be applied to. This isn't usually the case with most colocated racks or with most server rooms, in my experience. * "Do things right the first time so you don't have to do them again". This applies generally to things like figuring out which cable goes where. "Right" is not necessarily the most aesthetic or "neat" option, but it is the laziest and most time-efficient, with an eye for long-term maintenance.
The techniques are:
For ethernet: * ethernet should not be bundled approximately 5 per group. * ethernet cable should be a rainbow of color, with different colors in each bundle. Eg: a bundle should have blue, green, red, white, etc. not multiples of any given color. This helps drastically when determining which cable goes to which system. * Jack ends should be labeled descriptively and dated at the time of labeling. The description should describe the other end of the cable, not the end you're labeling (eg: sw2j5 for the 2nd NIC in server 5 that goes to switch 2 jack 5, or 'svr5n2' on the other end). * use generic (not the 'fancy' ones) velcro straps to bundle the cable * do not bundle the bundles, especially with velcro (because it will stick to the bundle velcro and make things a mess). To hold them up and away, I prefer using 6" ball bungees (http://www.amazon.com/Ball-Bungee-Cord-6-White/dp/B000S5TWWO) or similar, * DO NOT USE HORIZONTAL CABLE MANAGEMENT PRODUCTS. DO NOT. They make a mess of things and, more often than not, get in the way more often than not. The one exception to this is on a part of the rack where you've got a patch panel or switching which is likely to remain consistent through several server iterations. * Do not pull the Ethernet to one side of the rack only. "Alternate" the bundles of ethernet to each side (like pigtails, sorta).
For power: * Do not use the 6'+ power cables unless you need to. The cables are big enough and cheap enough to buy them in shorter lengths so you do not need to bundle them as much and the bundles are neater. * Label both ends with the name of the system in question (UPS side) and the UPS in question (server side). * Color code, if possible, by UPS. IE, if you have 2 PSUs per host, the first PSU gets black cables which go to one UPS, and the second gets grey cables which go to the other.
Wow, that really does sound dangerous. I suppose the WVa murder rate is higher than in most other places in the country, mainly the clean(er) city centers, then?
Who the hell would say you can "only use 400Mbit" of a gigE connection?
The only possible context I can put that in is Windows file sharing (where it's limited by shit protocol implementation/overhead). I, personally, regularly pull/push 90MB/s -105MB/s (that's megabyte) over NFS amongst multiple systems. Even on samba, pulling 75MB/s is tenable (though a bit difficult unless it's just Linux).
Honestly, I'm not sure how wireless or 'the average machine' even comes into question.
You can easily saturate gigE with the 'average' computer. I've done it with a 4-year-old Windows 7 laptop, no problem. Now, writing to storage at gigE speeds on the average machine is another question, but it's certainly tenable to push data at those rates. (He's probably basing his claims on shit software/driver implementations on Windows which impede throughput, anyway.)
His argument foregoes the three primary reasons why someone would legitimately want a relationally "fast" internet connection. On it's face, his argument is stupid.
* Faster porn downloads * Faster software downloads * Hosting your own porn/etc. site (the primary means through which "Profit!!!" is realized online)
I mean, seriously. What geek hasn't at least thought, "Hmm, for that extra $20-50 a month on what I'm paying now I can host services/servers for profit and quit my shit job"? I know people that have moved to a more expensive, smaller rental to get better fiber-based bandwidth. I mean, there are still businesses paying a shitload more than $100/month for T1 speed (and sometimes T1 service is, in fact, not much better than home cable/DSL on any given metric).
I did that as well, though in order to help it retain austerity in the conversation, I did a couple of things.
First, I did a database pre-seed from existing chat conversations from IRC, mimicking (in grammar choice) a handful of chatters I was familiar with to give it its personality.
Topically, it wasn't allowed to 'learn' - eg. factoids. Part of what i did this was to have the bot ignore assignments in person's conversations as a preliminary filter.
It's knowledge was defined by relational sets of topical words. On just a handful of topics, it was "knowledgeable" (eg. small talk). This information was partially pre-seeded and partially 'live' (I had it gather information from google, wikipedia, thesaurus.com, and various other sources as-needed to have coherent responses) . The portrayal (actual sentences) of any given 'fact' was also largely determined by variable synonyms and sentence structures.
Finally, the other thing I did was cause for a little human-like variance on things like punctuation, typos, capitalization, and spelling (different than typos). It was able to start tangental and follow-up conversations and track individual users, remembering what was said for each user (albeit, this part did not work as well as I'd have liked).
The only way I was able to retain some degree of "that may be a person" was to have the "randomly flighty/carefree and intellectually uninterested person" capability. Posing the bot as a younger 'cute' female on IRC helped, but so did carefully picking the channels I put it in combined with knowing what kind of words to have it trigger "flirt" and "insult" type responses.
Judging by the Indian programmers and support I've had the displeasure of working with, the logical divorce/completely non sequitur responses of this bot likely seem quite human to them - language capabilities aside!
Here, try this (exceedingly generic) 'emotional' conversation out that I just had with this thing:
Me: How you doing? Bot: I'm doing fine. Me: It's been a while. Bot: I'm glad. Me: What? you're glad? Bot: Because you're happy. Me: I'm not happy! you said you were glad we hadn't seen each other in a while! Bot: No, I said I was a girl. You were the one who said I was a boy.
There were better bots than this 20 years ago. This doesn't even approach '14 year old valley girl' intelligence.
The only thing this bot appears to be good at is interjecting non-related information into the 'conversation' and pausing variably between responses. A bot I wrote around 2003 did this as well, but it had per-user history analysis, deterministic responses, no pre-constructed sentences (relying upon a database of word relationships and sentence structure) and was able to maintain a conversation for about 20 private exchanges on average on IRC. Most of the time, there was no indication that the chatter realized it was a bot, and I had it sit in channels for weeks chatting people up.
If I could do it in a couple weeks of off-time programming, I'm sure there are many other examples of something drastically superior. This is not one of them.
Slashdot Mirror
You must not be playing anything all that great, then. Anything with 5.1 or even sometimes 2.1 surround typically fails, higher bitrate stuff is often choppy, and often .mkv stuff won't play properly or allow you to pause/resume/seek. This is with the built-in video stuff as well as players like RockPlayer or MoboPlayer.
Microsoft's intent has been for sometime to remove their dependence on other vendors. What makes you think anything more than a thin client will be needed in a couple years? That is quite obviously the direct MS is going with Windows 8. Windows 8 + an ARM thinclient + MS terminal service applications. All the better if they're Office 365 applications, right?
You're correct; however, if I can detect someone's 2.4GHz wireless security camera(s), it's another story. It really would be an invasion of privacy at that point, assuming I'm intent upon invading your privacy.
Had I known this 10 years ago when I started going to school to get my degree, I'd have gone into pre-law, and be able to retire in just a couple years after a successful career in software patent litigation...
Yeah, for infrastructure/interconnects/uplinks I'll use pink or red, and keep that out of my 'bundling', if it makes sense for the environment. But otherwise, it's a grab bag of different colors.
The labeling isn't there to make tracing the cable easy, it's there to keep track of which cable end is plugged into which port on which device.
No, it does both. The label assists with tracing, which labeling the 'correct' side doesn't do (telling you where that cable is plugged in on the other side). It's the only thing you can't tell by labeling the side 'correctly'.
Meanwhile, you can still look at a cable that's plugged into a device and see "oh, hey, this cable is plugged into the device". Labeling it with that device's name saves you no time, and a situation where a cable would be unplugged on one end would be fairly easy to detect from either side. (Labeling the opposite end tells you where you need to look to figure out which machine is unplugged, however.)
Labeling 'correctly' serves no better function, while delivering less.
I'll add to this.
First, clean the drive.
During the later days of floppy drives, the quality of the drives themselves went down. I'm not sure what it was, but they were less able to read and write floppies than the older drives. I still have some 3.5" floppy drives sitting in my 'recovery' (along with a board/etc. that can use it) which served me well. TEAC brand, IIRC. Both came out of ancient Gateway 2000 machines (486 and/or early Pentium towers, probably).
Don't use the USB/floppy converters/drives. They're almost worthless. Once floppies got around to about $0.50/ea, the drives themselves were nearly as useless.
From my recollection, dd on linux would seek until it got a good reading. More often than not, most of what was recovered was usable even with that. Failing that, there's also ddrescue and the other tools being recommended. (Linux also seems more able to read floppies than Windows or Mac, for what it's worth. I was usually able to 'recover' people's data from the disk simply by copying it to hard drive and re-writing it to another floppy.)
I would argue that there's really no good reason why the later android builds can't be made to work on the older phones. There are a number of Android phones with some fairly minimal specifications (500MHz CPU, 128-256Mb of RAM) which are quite responsive with the 2.3.x android build, and there are a number of phones that perform poorly simply due to the specific software build, not necessarily the "newness" or "featureful" nature of the software.
As an HTC HD2 owner, having tried a number of different ROMs, I've noticed that there is a huge variance in performance even on the same basic kernel/android version due to how the ROM was set up - everything from the specific 'tweaks' on how dalvik/caching/CPU is configured to which managers and widgets are used.
Oh really?
Just off the top of my head:
WP7:
* primarily uses Exchange, which is (in its online/prevailant incarnation of bpos/office365) incompatible with anything but MS clients
* has a mobile MS Office built-in (some call that a feature, and I'll admit, OneNote is sweet.)
* does not release their source code
Android:
* can use a myriad of mail services, including gmail, from which I can export all my mail from any client, of which there are several. I can also export my contacts in any format I please (literally)
* has a number of available word processor type things.
* releases their source code (for the most recent phone release, which I think is 2.3.4 at this point) for everything not dealing directly with their infrastructure
* allows 3rd party software to be installed w/o boostrapping to their market (hmm, you weren't thinking winmo6.5 were you?)
This is simple to explain away.
Unlike "before", when there were some fairly interesting and innovative ideas on the mobile realm from open source in both hardware and software, there's actually a fair amount of R&D competition. Nokia's N800, N880, N900 and such, on Maemo, were open source to their core, with some fairly solid design philosophies. That was about it in terms of "open source mobile", and it had a fair amount of effort behind it (it, as well as fltk, tinyX, etc. on which it was based, and then later, Qt).
Now, however, you can actually get a job to work on this shit. Once taht's done, however, you're likely to have an NDA/noncomp which would inhibit your ability to work on OSS... oh well. (They're likely making serious bank in a down economy; don't jinx that shit.)
No kidding. They should just be happy to humiliate women for the humiliating treatment of women. It would be greatly preferable to the joke that the refer to as "ethical" treatment of animals.
(I miss being able to afford fur coats. Anyone remember back when you could actually buy a mink, fox, or coyote jacket for the winter without importing it from Russia or China?)
Your last point is easily mitigated by taking your second to last point a step further:
Lastly, limit access to the wiring closets only to those that need it, have been trained, and are held responsible if it becomes a mess.
You can use different cable colors for identifying certain things in your environment (wireless, printers, servers, etc). If you can't justify buying all of the cables sizes you need in all of the different colors then you can use colored tape or some other type of identifier like plastic tags. You have plenty of options.
What I propose and do myself is: don't care what color the cables are, as long as they're different. Then label them. Or, don't label them, but having different color cables in each bundle makes the "figure out where cables go" problem so very much easier and almost completely avoidable. You still have to trace cables with colored cables, and when they're colored by role, you'll often have bundles of same-colored cables. That's no good. Just mix and match, who gives a damn (the different colors per bundle make it easier to find and identify bundles as unique, too). You don't need to 'train' anyone to your specific environment for little things, and it makes it much easier to hand off to someone else.
At first I thought you were being sarcastic with a statement like this:
Proper use of cable ties means you are not afraid to use a LOT of them, and also not afraid of cutting them open when you need to change someting. I keep a cheap diagonal cutter and a bag of assorted cable ties in every desk drawer in my house (3 "kits" in total).
That makes sense in some applications, so maybe they do for your's. You weren't really specific.
However, let me just say that as someone who has come after the dipshits who cinch cable tied an entire rack of equipment in (and the people who came after them, who tried to make due with having all their cables constrained and inaccessible), this is a bad practice. Not only does it result in marginal forethought, but it requires two things you do not want:
1) You need an extra tool (and more time) to redo wiring or even rack another machine, sometimes.
2) That you have a knife near networking cable which is able to cut through the ties - ties which are more difficult to cut than the networking cable itself.
Case in point for #2: we were reorganizing a rack (removing every third/other system, moving the remaining machines down, and putting new equipment in the top). We were almost done and checking to make sure we didn't jiggle something loose or fail to plug something back in properly, and we noticed that the main DC was intermittently available. Someone had managed to nick the cable in a bundle when separating it and didn't notice - slicing through the cable enough to cause some fairly severe (but not complete) packet loss.
The ONLY time I would say a network/server rack should be cinch tied is when you are building a handful of racks to a universal specification and will not be touching the cable until the cable or equipment is completely replaced.
Do not use this shit - either the cable wrap or the cinch ties - in a server room. Please.
It may look nice, and for applications where "this looks really neatly done", that's important. A server room or a networking rack is not one of these places. In such places, it will only get in the way, turning a quick job into a lengthy one.
Like any addiction, the first step is to admit that you've got a problem. :)
Seriously, though. You've got to realize that whatever the result of your efforts is, it's still going to be non-ideal, even if it looks as uniform and as neat as humanly possible. The truth is, no matter of initial neatness will prevent someone from coming along and fucking it up, whether unintentionally and with good cause or because they're a lazy git.
At home, I will label both ends of a cable that is likely to be sitting in a pile with other cables (usually - particularly when they're likely to be confused with others and when tracing the cable will be difficult). Just a simple file folder label or piece of tape, usually. Mostly, I don't worry about it, because at home "oops, I unplugged the system" only bothers me and/or my family. At work, it's another story: everything you do should be done to help mitigate downtime and improve your ability to effectively work with the cabling.
On server racks, I prefer a handful of techniques. There are a couple principles I abide by:
* always assume the rack will get messy over time
* your original intent will not be the intent of the next person to come along
* never assume a standard, because everyone has a different idea of how things should be done, and first impressions to that effect can be wrong
* standards only make sense when the implementation requirements/specifications/etc. are identical/universal, eg. with electrical wiring or in a large hosting facility where everything is the same or there is a standard which can be applied to. This isn't usually the case with most colocated racks or with most server rooms, in my experience.
* "Do things right the first time so you don't have to do them again". This applies generally to things like figuring out which cable goes where. "Right" is not necessarily the most aesthetic or "neat" option, but it is the laziest and most time-efficient, with an eye for long-term maintenance.
The techniques are:
For ethernet:
* ethernet should not be bundled approximately 5 per group.
* ethernet cable should be a rainbow of color, with different colors in each bundle. Eg: a bundle should have blue, green, red, white, etc. not multiples of any given color. This helps drastically when determining which cable goes to which system.
* Jack ends should be labeled descriptively and dated at the time of labeling. The description should describe the other end of the cable, not the end you're labeling (eg: sw2j5 for the 2nd NIC in server 5 that goes to switch 2 jack 5, or 'svr5n2' on the other end).
* use generic (not the 'fancy' ones) velcro straps to bundle the cable
* do not bundle the bundles, especially with velcro (because it will stick to the bundle velcro and make things a mess). To hold them up and away, I prefer using 6" ball bungees (http://www.amazon.com/Ball-Bungee-Cord-6-White/dp/B000S5TWWO) or similar,
* DO NOT USE HORIZONTAL CABLE MANAGEMENT PRODUCTS. DO NOT. They make a mess of things and, more often than not, get in the way more often than not. The one exception to this is on a part of the rack where you've got a patch panel or switching which is likely to remain consistent through several server iterations.
* Do not pull the Ethernet to one side of the rack only. "Alternate" the bundles of ethernet to each side (like pigtails, sorta).
For power:
* Do not use the 6'+ power cables unless you need to. The cables are big enough and cheap enough to buy them in shorter lengths so you do not need to bundle them as much and the bundles are neater.
* Label both ends with the name of the system in question (UPS side) and the UPS in question (server side).
* Color code, if possible, by UPS. IE, if you have 2 PSUs per host, the first PSU gets black cables which go to one UPS, and the second gets grey cables which go to the other.
... you DO know they're talking about the Windows 8 interface, right? Have you seen it? It makes the Apple interface look, like, totally boring.
Teach them to 'hum' Flight of the Valkyries - badly. If i were a bird, and didn't wear pants, that'd be what I'd sing.
Wow, that really does sound dangerous. I suppose the WVa murder rate is higher than in most other places in the country, mainly the clean(er) city centers, then?
I would kiss you if I could. Hilarious response.
Who the hell would say you can "only use 400Mbit" of a gigE connection?
The only possible context I can put that in is Windows file sharing (where it's limited by shit protocol implementation/overhead). I, personally, regularly pull/push 90MB/s -105MB/s (that's megabyte) over NFS amongst multiple systems. Even on samba, pulling 75MB/s is tenable (though a bit difficult unless it's just Linux).
Honestly, I'm not sure how wireless or 'the average machine' even comes into question.
You can easily saturate gigE with the 'average' computer. I've done it with a 4-year-old Windows 7 laptop, no problem. Now, writing to storage at gigE speeds on the average machine is another question, but it's certainly tenable to push data at those rates. (He's probably basing his claims on shit software/driver implementations on Windows which impede throughput, anyway.)
His argument foregoes the three primary reasons why someone would legitimately want a relationally "fast" internet connection. On it's face, his argument is stupid.
* Faster porn downloads
* Faster software downloads
* Hosting your own porn/etc. site (the primary means through which "Profit!!!" is realized online)
I mean, seriously. What geek hasn't at least thought, "Hmm, for that extra $20-50 a month on what I'm paying now I can host services/servers for profit and quit my shit job"? I know people that have moved to a more expensive, smaller rental to get better fiber-based bandwidth. I mean, there are still businesses paying a shitload more than $100/month for T1 speed (and sometimes T1 service is, in fact, not much better than home cable/DSL on any given metric).
I did that as well, though in order to help it retain austerity in the conversation, I did a couple of things.
First, I did a database pre-seed from existing chat conversations from IRC, mimicking (in grammar choice) a handful of chatters I was familiar with to give it its personality.
Topically, it wasn't allowed to 'learn' - eg. factoids. Part of what i did this was to have the bot ignore assignments in person's conversations as a preliminary filter.
It's knowledge was defined by relational sets of topical words. On just a handful of topics, it was "knowledgeable" (eg. small talk). This information was partially pre-seeded and partially 'live' (I had it gather information from google, wikipedia, thesaurus.com, and various other sources as-needed to have coherent responses) . The portrayal (actual sentences) of any given 'fact' was also largely determined by variable synonyms and sentence structures.
Finally, the other thing I did was cause for a little human-like variance on things like punctuation, typos, capitalization, and spelling (different than typos). It was able to start tangental and follow-up conversations and track individual users, remembering what was said for each user (albeit, this part did not work as well as I'd have liked).
The only way I was able to retain some degree of "that may be a person" was to have the "randomly flighty/carefree and intellectually uninterested person" capability. Posing the bot as a younger 'cute' female on IRC helped, but so did carefully picking the channels I put it in combined with knowing what kind of words to have it trigger "flirt" and "insult" type responses.
Judging by the Indian programmers and support I've had the displeasure of working with, the logical divorce/completely non sequitur responses of this bot likely seem quite human to them - language capabilities aside!
Here, try this (exceedingly generic) 'emotional' conversation out that I just had with this thing:
Me: How you doing?
Bot: I'm doing fine.
Me: It's been a while.
Bot: I'm glad.
Me: What? you're glad?
Bot: Because you're happy.
Me: I'm not happy! you said you were glad we hadn't seen each other in a while!
Bot: No, I said I was a girl. You were the one who said I was a boy.
There were better bots than this 20 years ago. This doesn't even approach '14 year old valley girl' intelligence.
The only thing this bot appears to be good at is interjecting non-related information into the 'conversation' and pausing variably between responses. A bot I wrote around 2003 did this as well, but it had per-user history analysis, deterministic responses, no pre-constructed sentences (relying upon a database of word relationships and sentence structure) and was able to maintain a conversation for about 20 private exchanges on average on IRC. Most of the time, there was no indication that the chatter realized it was a bot, and I had it sit in channels for weeks chatting people up.
If I could do it in a couple weeks of off-time programming, I'm sure there are many other examples of something drastically superior. This is not one of them.