He goes on to say that 'Linux is not ready for mission-critical computing.
In general, I agree with him on this (I have not RTFA yet). Nor is Windows, of course, but that's taken for granted. Of course, it depends how critical your mission is. "Mission-Critical" is one of these phrases which is bandied around, but let's consider what it means....
"The mission depends on this system".
That still does not define the extent to which the mission depends on it - 80%? 90%? 100%? Nobody offers 100% availability, if that's what you're referring to.
The phrase also ignores the mission involved. For NASA, the Mission might be to send a man to Mars and back, but what if my "mission" is to run a website which expects to get 3 hits a month with a 60% expectation of success? An Atari could cope with that - my mobile phone could probably cope with that!
Taking the phrase in the way it's normally meant (running systems which are responsible for a significant amount of the user's business, and the failure of which would cause significant disruption of the business process and/or profit), then the whole discussion still depends entirely on the "mission" involved.
What tradeoffs is the mission prepared to make for uptime, for example? Serving read-only webpages, I care little for data integrity (I've been serving the same data for years, I've got it on tape, CD, DVD, onsite and offsite), and only care about uptime.
If I'm running a database which is updated many times a minute, then uptime still matters to me, but I also need to know which transactions have been fully processed, and which have failed (given Failure Scenario N, which may or may not have been predictable). That is much more difficult.
It's not likely. Could be btinternet.com, could be btopenworld.com, or btconnect.com; when I use the dialup service provided with my BT ADSL account, I get redirected to a page on business.btconnect.com (or something like that) which (naturally) doesn't recognise my login credentials, as I'm a residential, not a business, customer!
Eedjits.
Interestingly, I phoned them last week to get a MAC code (necessary to migrate an ADSL connection) and - instead of giving me that, they asked why I wanted to move (download caps coming soon at BT) and they gave me a 512k->1Mbps upgrade, with a 30Gb/month (instead of the 1Gb/month) cap when they do bring it in. That'll do me, but if they're happy to do that (for no extra charge), why do they try to enforce these limitations for those who don't threaten to leave?
On the plus side (OMG, I'm saying something good about BT!), back in the days before they outsourced their mail hosting to Yahoo!, they used Exim (www.exim.org) as their mail server.
I noticed this shortly after switching to Exim for our own mail server... gave me a few twitchy feelings, but Exim's been doing us good service for a few years now.
I don't know, but it's much more like that he "wore a leather jacket" than that he "weared a leather jacket" - unless he took the jacket into the bank, where he scrubbed it briskly with a steel brush.
Heh - tell me about it, I've just got back to England from a few days in Glasgow, and realised that I've still got a Scottish fiver in my pocket (despite my best attempts to spend all "funny money" while north of the border!)
I can see that small shopkeepers find it difficult to train their staff to check English money carefully, let alone Scottish money, too, but it's all legal tender; what next? "Can't take that Mint card, it's got a bit missing off the corner"?!
It probably comes under "Fair Use", but a very good point all the same - the whole F/OSS movement is very strongly based on correct attribution, and just as the code has to be developed carefully, it wouldn't hurt if the most prominent website about F/OSS took just a little care about proper attribution.
Better still, (no, please, don't laugh, just hear me out...) Slashdot should be setting an example - raising the bar of properly accredited journalism.
Okay, you can laugh now. I know you've been holding it in.
Getting home in time to see the kids is the best part of my day. When I'm away from home, or just too late home, I miss out on another day of their lives.
That's a big hit. Okay, I got Customer X's network running smoothly, but little B and baby E didn't see me that day - and I didn't see them. As a one-off, that's okay. Day-to-day, I don't see my role of father as "leave home before the kids are up, get back after they're in bed; maybe see them for a while on the weekend." That is not parenthood.
A job I was in recently is a good example - a 2h commute each way, and they said, "12h is seen as normal; 16h isn't at all unexpected." Fsck that. 12h + 4h = 16h (8h sleep and back to the job); 16 + 4h leaves me 4h to sleep!
The question is: "Do I live to work, or do I work to live?"
I work to live - my work is interesting enough - could be better. It brings in cash, and supports my *real* life - my family, my kids; if I don't spend time with my wife and kids, then what is the point?! They could live a grand life, in a huge house, driving around in grand cars, but simply fail to recognise their own father. That is not a life for them, or for me.
I have 100% admiration for single parents, but this way of life just gives the family a single-parent-family lifestyle, with a breadwinner who - technically - exists, but who is never seen. What's the point in that?
I'd rather bring in enough cash to keep the family on track, with enough time to actually spend time with them - teaching the kids, and enjoying them. Spending time with my wife is also a priority, of course - after all, that is why I married her - I love her!
So you spend your time writing English, not C... but there are two v. simple grammatical errors in this post:
Its very nerve racking
You mean "It's," not "Its", and you should have said "nerve-wracking", not "nerve racking".
but somehow there careers have migrated
You meant to say "their," not "there."
Okay, so I'm being pedantic, but you did post on/., and if you're going to go on about the wonders of speaking to people - as opposed to speaking to machines - then you should at least show the same grammatical understanding that a programmer has.
Call me a grammar nazi if you like - feel free, in fact - but I am sick to the hind teeth of people (technical, non-technical, anyone) claiming to speak English without actually doing so.
I just don't understand this IBM-cheerleading which/. is so full of. So the BigCo is being sued by SmallCo, and we don't like SmallCo's attitude.
Is that world that black-and-white, that we forget everything that BigCo has done? Like, using its patents to kill off competitors, for example?
Okay, IBM was a sucker to MS, way back, but that does not make IBM the underdog. Are you running IBM-Linux? No, nor am I. They'll sell you a flavour of Linux, if you decide you want RedHat or SuSE. They don't really care - you're unlikely to want AIX, so RedHat/SuSE is cheaper for IBM to resell than Windows is. Most of their customers get Windows, though. Reselling 3rd-party Linux distro's is just a (slightly) cheaper way for IBM to sell tin.
What is this compulsion among slashdot that IBM actually contribute anything useful to Linux?
Maybe part of it is the confusion about what "Linux" is, and part of that, is the refusal to accept there is more GNU code than Linux code in a typical distribution. Again, the black-and-white, "Linus is Good, RMS is on an ego trip" attitude.
Go for it - run Linux, without gnu's ls. Install IBM-Linux onto your kit, once it's available.
It's not got to the "emperor has no clothes" stage yet, but it's not far off, is it?
IBM shifts tin; they've presumably contributed to Windows code - at least in feedback, if not in actual lines of C, just as they have with Linux, and lots of surrounding projects. So what?
They are not the underdog, they're just another hardware manufacturer. I don't get it.
PLEASE PLEASE PLEASE ENLIGHTEN ME WHAT IBM HAVE DONE FOR "LINUX" (however you define that) MORE THAN ANYONE ELSE.
I can't see this as a "community" thing any more - I used to, but the more it goes on, it's just some contract deal between IBM and USL-or-whoever-they-signed-the-paperwork-with (I forget). Nothing to see here. Move along.
Because UNIX was (is) proprietary code, and, while everyone was sharing it pretty freely, the lawyers had their hands in the pot. Everybody understood that it was a gentleman's agreement, and one gets the impression that nobody looked too closely at the small print, because everyone was in the same bubble of mutual dependancy, we're sharing it all anyway, and even if we don't, then there's BSD over there with the same stuff "for free" so the documentation is just a bit of paperwork.
We won't go around publishing the lot for free in an ACM newsletter, but we all know that all UNIX licensees have access to the stuff anyway.
That seems to be the spirit in which the original licenses were signed; When Sun changed from BSD to SysV, there was a lot of fuss from customers, but nobody seemed to worry about "but now we're beholden to AT&T/USL" - it wasn't an issue.
Everybody was in the same clique.
Now, new-SCO seem to be re-interpreting old documents, but rather too late, IMNSVHO. Particularly now that everything (SysV, Linux, even Windows), has tons of BSD code in it, the original license is a historic document, at best. The more obscure parts, like "what you derive, you give back (which is rather GPL-ish, if you think about it, but without the "give back to everyone" aspect), have been declared void by previous holders of the license (USL/Novell), but SCO are trying to revive that aspect of it. Good luck, SCO. The world is against you.
Because, according to their theory, the same stuff can be found on kernel.org (which is just as accessible from internal.sco.com as from anywhere else).
They don't need to hand over "their" code - they're not talking about anything they've written in-house, they're saying that IBM wrote stuff for SysV, and then put it into Linux. SCO don't know what IBM put into AIX and/or Dynix any more than you or I know. They think the license they bought from USL entitles them to the code, though.
I'd have thought that the sensible thing to do would be to work out if SCO's argument holds true before delving into the code - after all, courts are much better at dealing with details in contracts than they are at working out what bits of C code look like other bits of C code....
See Sunsolve. The IEEE specs are open to various interpretations; this can lead to Gb interfaces going to 100/hdx or other dodgy configs. See also Cisco's website for their take. (Also see here.)
Cisco seem to recommend autonegotiation; Sun recommend forcing the speed/duplex.
We've had problems in the past with Sun's "ce" fibre cards and Cisco Catalyst switches. It's not that either implementation is "wrong", the specs simply are not specific enough.
Sorry, can't find the detail in the spec which causes the problem
I'm pleased to say that I've never (AFAIK) been the culprit, but I've been around for a few goodies - two being the classic "but I thought that was the door-release". One of these just hit the Big Red Button as someone happened to be entering, so the door opened, and the culprit wandered out of the machine room without noticing that it all went dark and quiet behind him.
The second was a guy who was on his first day of work with us. A Big Boss came towards the machine room, so - feeling helpful - the new guy opens the door for him... or so he thought.
My favourite story (though I wasn't there) is about some old DEC machines, which apparently had the power switch about 6" from the floor. Nobody knew why they kept crashing at night, until someone spotted a cleaner ramming a vacuum cleaner right up to the servers.
That beats the one we had, when I used to do a lot of soak-testing of machines in a lab - I'd kick off a test on a Friday night, come back in on Monday to find the machine had rebooted. Nothing in the logs, just looked like the power had died, and returned again half an hour later. Other machines on the same power supply were fine.
It turned out that the cleaners were unplugging the servers, so they could plug in the vacuum cleaner!
Bullshit.
Speed/Duplex negotiation is an OS configuration issue, not a hardware NIC issue.
If the OS can't configure the negotiation, that's still the OS, not the hardware. It just means that the driver isn't capable of properly configuring the NIC. Just because your workaround was in hardware, does not mean that that is the cause of the problem.
As for adding your own UPSes which ignore the EPO, is surely that defeats the object of the EPO. I don't know USAian requirements, but if, as you say, the EPO is required, is it legal to bypass it with your own UPSes?
LJ clearly have not heard of DR; although a true DR configuration is probably overkill for this type of site, this report gives the strong impression that basic sysadmin competencies were not followed when there was time available - during design and deployment, and then later during normal running. These problems had apparently not occured to anyone until it happened. Isn't "what's the worst-case scenario" a common-enough question? Wouldn't "total power failure" be one of those answers?
Even with write-though caches, a small battery in the array can flush data to disk after a power failure. This isn't rocket science - buy the right kit for the job, understand what you're buying, and how to configure it. If you don't understand what it is, what on earth made you decide to buy it?!! You've got dual-powered systems, but didn't use that feature - why did you buy it then? It wasn't a conscious decision to take the risk, and it wasn't a conscious decision to get dual-powered hardware for resilience. No thought was made about power. Most colo's provide dual-sourced power supplies for this type of problem - power from seperate grids, so even if the grid providing power to the datacentre goes down, the alternate grid continues running.
Sigh. I often have customers nearly as daft as this, though I don't think I've come across such a poorly considered deployment for a long, long time.
The engineer shoud be paid enough (and educated enough) to tell their employer, "If I make it available to you, it's available to the internet,and (if we're in a vaguely rational county (eg, UK)) that would be illegal." So - No, I won't do that.
It's all difficult to read, given the bad English, the America-centrisim, and poor arguments, but the article as stands, if rather sort, doesn't seem too inflammatory.
I've been given all sorts of definitions of a "null-modem" - I'd rather go to www.nullmodem.com than a general-purpose database (wikipedia is, after all, just a front-end to a database).
As for "what does ecology mean?," you'll never get that answer from a single web page, it's a discussion, which does not have a single "correct" answer.
You seem to have supplied an excellent example on the weaknesses of Wikipedia.
When it started out, I expected it to die within a year It didn't. I was wrong.
I never refer to it myself, though, because I do not trust it. How could I - even if a friend referred me to your article, I can't be sure of the source when I get around to viewing it.
Who told you that that was the original intent of the Internet? They were wrong. The internet is a method of transmitting data. It was created for military and academic sharing of information, but that sharing was fully accredited, not some random information supplied by John and Jane Doe (no address supplied). It was intended for qualified professionals to publish information - editing came much, much, much later (even later than the web)
Thank you for translating the random words into English! Yet another example of why education helps.
(Quick plug for "Lost for Words" by John Humphrys)
Slashdot Mirror
In general, I agree with him on this (I have not RTFA yet). Nor is Windows, of course, but that's taken for granted. Of course, it depends how critical your mission is. "Mission-Critical" is one of these phrases which is bandied around, but let's consider what it means....
"The mission depends on this system".
That still does not define the extent to which the mission depends on it - 80%? 90%? 100%? Nobody offers 100% availability, if that's what you're referring to.
The phrase also ignores the mission involved. For NASA, the Mission might be to send a man to Mars and back, but what if my "mission" is to run a website which expects to get 3 hits a month with a 60% expectation of success? An Atari could cope with that - my mobile phone could probably cope with that!
Taking the phrase in the way it's normally meant (running systems which are responsible for a significant amount of the user's business, and the failure of which would cause significant disruption of the business process and/or profit), then the whole discussion still depends entirely on the "mission" involved.
What tradeoffs is the mission prepared to make for uptime, for example? Serving read-only webpages, I care little for data integrity (I've been serving the same data for years, I've got it on tape, CD, DVD, onsite and offsite), and only care about uptime.
If I'm running a database which is updated many times a minute, then uptime still matters to me, but I also need to know which transactions have been fully processed, and which have failed (given Failure Scenario N, which may or may not have been predictable). That is much more difficult.
You did check the URL first, right?
Eedjits.
Interestingly, I phoned them last week to get a MAC code (necessary to migrate an ADSL connection) and - instead of giving me that, they asked why I wanted to move (download caps coming soon at BT) and they gave me a 512k->1Mbps upgrade, with a 30Gb/month (instead of the 1Gb/month) cap when they do bring it in. That'll do me, but if they're happy to do that (for no extra charge), why do they try to enforce these limitations for those who don't threaten to leave?
I noticed this shortly after switching to Exim for our own mail server... gave me a few twitchy feelings, but Exim's been doing us good service for a few years now.
I don't know, but it's much more like that he "wore a leather jacket" than that he "weared a leather jacket" - unless he took the jacket into the bank, where he scrubbed it briskly with a steel brush.
I can see that small shopkeepers find it difficult to train their staff to check English money carefully, let alone Scottish money, too, but it's all legal tender; what next? "Can't take that Mint card, it's got a bit missing off the corner"?!
I'll take one of those dual-process 8086 machines, if you've got a spare one :-)
It probably comes under "Fair Use", but a very good point all the same - the whole F/OSS movement is very strongly based on correct attribution, and just as the code has to be developed carefully, it wouldn't hurt if the most prominent website about F/OSS took just a little care about proper attribution.
Better still, (no, please, don't laugh, just hear me out...) Slashdot should be setting an example - raising the bar of properly accredited journalism.
Okay, you can laugh now. I know you've been holding it in.
That's a big hit. Okay, I got Customer X's network running smoothly, but little B and baby E didn't see me that day - and I didn't see them. As a one-off, that's okay. Day-to-day, I don't see my role of father as "leave home before the kids are up, get back after they're in bed; maybe see them for a while on the weekend." That is not parenthood.
A job I was in recently is a good example - a 2h commute each way, and they said, "12h is seen as normal; 16h isn't at all unexpected." Fsck that. 12h + 4h = 16h (8h sleep and back to the job); 16 + 4h leaves me 4h to sleep!
The question is: "Do I live to work, or do I work to live?"
I work to live - my work is interesting enough - could be better. It brings in cash, and supports my *real* life - my family, my kids; if I don't spend time with my wife and kids, then what is the point?! They could live a grand life, in a huge house, driving around in grand cars, but simply fail to recognise their own father. That is not a life for them, or for me.
I have 100% admiration for single parents, but this way of life just gives the family a single-parent-family lifestyle, with a breadwinner who - technically - exists, but who is never seen. What's the point in that?
I'd rather bring in enough cash to keep the family on track, with enough time to actually spend time with them - teaching the kids, and enjoying them. Spending time with my wife is also a priority, of course - after all, that is why I married her - I love her!
Okay, so I'm being pedantic, but you did post on /., and if you're going to go on about the wonders of speaking to people - as opposed to speaking to machines - then you should at least show the same grammatical understanding that a programmer has.
Call me a grammar nazi if you like - feel free, in fact - but I am sick to the hind teeth of people (technical, non-technical, anyone) claiming to speak English without actually doing so.
I just don't understand this IBM-cheerleading which /. is so full of. So the BigCo is being sued by SmallCo, and we don't like SmallCo's attitude.
Is that world that black-and-white, that we forget everything that BigCo has done? Like, using its patents to kill off competitors, for example?
Okay, IBM was a sucker to MS, way back, but that does not make IBM the underdog. Are you running IBM-Linux? No, nor am I. They'll sell you a flavour of Linux, if you decide you want RedHat or SuSE. They don't really care - you're unlikely to want AIX, so RedHat/SuSE is cheaper for IBM to resell than Windows is. Most of their customers get Windows, though. Reselling 3rd-party Linux distro's is just a (slightly) cheaper way for IBM to sell tin.
What is this compulsion among slashdot that IBM actually contribute anything useful to Linux?
Maybe part of it is the confusion about what "Linux" is, and part of that, is the refusal to accept there is more GNU code than Linux code in a typical distribution. Again, the black-and-white, "Linus is Good, RMS is on an ego trip" attitude.
Go for it - run Linux, without gnu's ls. Install IBM-Linux onto your kit, once it's available.
It's not got to the "emperor has no clothes" stage yet, but it's not far off, is it?
IBM shifts tin; they've presumably contributed to Windows code - at least in feedback, if not in actual lines of C, just as they have with Linux, and lots of surrounding projects. So what?
They are not the underdog, they're just another hardware manufacturer. I don't get it.
PLEASE PLEASE PLEASE ENLIGHTEN ME WHAT IBM HAVE DONE FOR "LINUX" (however you define that) MORE THAN ANYONE ELSE.
I can't see this as a "community" thing any more - I used to, but the more it goes on, it's just some contract deal between IBM and USL-or-whoever-they-signed-the-paperwork-with (I forget). Nothing to see here. Move along.
Because UNIX was (is) proprietary code, and, while everyone was sharing it pretty freely, the lawyers had their hands in the pot. Everybody understood that it was a gentleman's agreement, and one gets the impression that nobody looked too closely at the small print, because everyone was in the same bubble of mutual dependancy, we're sharing it all anyway, and even if we don't, then there's BSD over there with the same stuff "for free" so the documentation is just a bit of paperwork.
We won't go around publishing the lot for free in an ACM newsletter, but we all know that all UNIX licensees have access to the stuff anyway.
That seems to be the spirit in which the original licenses were signed; When Sun changed from BSD to SysV, there was a lot of fuss from customers, but nobody seemed to worry about "but now we're beholden to AT&T/USL" - it wasn't an issue.
Everybody was in the same clique.
Now, new-SCO seem to be re-interpreting old documents, but rather too late, IMNSVHO. Particularly now that everything (SysV, Linux, even Windows), has tons of BSD code in it, the original license is a historic document, at best. The more obscure parts, like "what you derive, you give back (which is rather GPL-ish, if you think about it, but without the "give back to everyone" aspect), have been declared void by previous holders of the license (USL/Novell), but SCO are trying to revive that aspect of it. Good luck, SCO. The world is against you.
They don't need to hand over "their" code - they're not talking about anything they've written in-house, they're saying that IBM wrote stuff for SysV, and then put it into Linux. SCO don't know what IBM put into AIX and/or Dynix any more than you or I know. They think the license they bought from USL entitles them to the code, though.
I'd have thought that the sensible thing to do would be to work out if SCO's argument holds true before delving into the code - after all, courts are much better at dealing with details in contracts than they are at working out what bits of C code look like other bits of C code....
Cisco seem to recommend autonegotiation; Sun recommend forcing the speed/duplex.
We've had problems in the past with Sun's "ce" fibre cards and Cisco Catalyst switches. It's not that either implementation is "wrong", the specs simply are not specific enough.
Sorry, can't find the detail in the spec which causes the problem
The second was a guy who was on his first day of work with us. A Big Boss came towards the machine room, so - feeling helpful - the new guy opens the door for him... or so he thought.
My favourite story (though I wasn't there) is about some old DEC machines, which apparently had the power switch about 6" from the floor. Nobody knew why they kept crashing at night, until someone spotted a cleaner ramming a vacuum cleaner right up to the servers.
That beats the one we had, when I used to do a lot of soak-testing of machines in a lab - I'd kick off a test on a Friday night, come back in on Monday to find the machine had rebooted. Nothing in the logs, just looked like the power had died, and returned again half an hour later. Other machines on the same power supply were fine.
It turned out that the cleaners were unplugging the servers, so they could plug in the vacuum cleaner!
Speed/Duplex negotiation is an OS configuration issue, not a hardware NIC issue.
If the OS can't configure the negotiation, that's still the OS, not the hardware. It just means that the driver isn't capable of properly configuring the NIC. Just because your workaround was in hardware, does not mean that that is the cause of the problem.
As for adding your own UPSes which ignore the EPO, is surely that defeats the object of the EPO. I don't know USAian requirements, but if, as you say, the EPO is required, is it legal to bypass it with your own UPSes?
LJ clearly have not heard of DR; although a true DR configuration is probably overkill for this type of site, this report gives the strong impression that basic sysadmin competencies were not followed when there was time available - during design and deployment, and then later during normal running. These problems had apparently not occured to anyone until it happened. Isn't "what's the worst-case scenario" a common-enough question? Wouldn't "total power failure" be one of those answers?
Even with write-though caches, a small battery in the array can flush data to disk after a power failure. This isn't rocket science - buy the right kit for the job, understand what you're buying, and how to configure it. If you don't understand what it is, what on earth made you decide to buy it?!! You've got dual-powered systems, but didn't use that feature - why did you buy it then? It wasn't a conscious decision to take the risk, and it wasn't a conscious decision to get dual-powered hardware for resilience. No thought was made about power. Most colo's provide dual-sourced power supplies for this type of problem - power from seperate grids, so even if the grid providing power to the datacentre goes down, the alternate grid continues running.
Sigh. I often have customers nearly as daft as this, though I don't think I've come across such a poorly considered deployment for a long, long time.
The engineer shoud be paid enough (and educated enough) to tell their employer, "If I make it available to you, it's available to the internet,and (if we're in a vaguely rational county (eg, UK)) that would be illegal." So - No, I won't do that.
So - since you seem to speak for WikiPedia - could you tell us what it is useful for?
It's all difficult to read, given the bad English, the America-centrisim, and poor arguments, but the article as stands, if rather sort, doesn't seem too inflammatory.
As for "what does ecology mean?," you'll never get that answer from a single web page, it's a discussion, which does not have a single "correct" answer.
You seem to have supplied an excellent example on the weaknesses of Wikipedia.
When it started out, I expected it to die within a year It didn't. I was wrong.
I never refer to it myself, though, because I do not trust it. How could I - even if a friend referred me to your article, I can't be sure of the source when I get around to viewing it.
Who told you that that was the original intent of the Internet? They were wrong. The internet is a method of transmitting data. It was created for military and academic sharing of information, but that sharing was fully accredited, not some random information supplied by John and Jane Doe (no address supplied). It was intended for qualified professionals to publish information - editing came much, much, much later (even later than the web)
Fix "occilating" to "oscillating"
I'd do it myself, but I trust nothing on Wikipedia, including spelling. If you care, then you fix it.
What a classic oxymoron!
Thank you for translating the random words into English! Yet another example of why education helps. (Quick plug for "Lost for Words" by John Humphrys)