No the internet is *not* working well these days.. it's working well until you piss off some kiddy who has a lot of waiting drones. Or until some worm hits.
And the problem is not one of archetecture it's one of lazy admins who can't be bothered to upgrade in 4 months and of management who demand applications based on vendors with poor security records.
I am no longer happy to have to make the needed phone calls when my stuff is being flooded offline. I'm at the point where I just want someone to give the legions of the lazy and incompetant to go off and find some other line of work where they don't have to affect the rest of us. If that means govt enforced fines or licensing.. so be it. It *has* to be better than what we have now.
Probably better odds elsewhere. After having actually seen what passes for security in the online gambling industry all you need really is a java decompilor.
Many of the online casino games tell the server whether it won or lost. And on one particularly funny case the game connected right to the SQL server at the casino.
The first place I worked did better with a flash/php combo but theve never bothered to upgrade apache or ssl since I quit.
It's quite sad really I'm supprised these places don't get ripped off more often.
Yeah well I too miss the days when a rooted server on someone else's network was not my problem. But welcome to today.
How exactly are network archetects supposed to design for 300 drones all sending traffic to one place? There is no amount of overcapacity that would handle that.
It's certainly not as bad as they are making out. It's only been a few months since one of worldcom's border routers went bad and took out several other isp's with it. My uplink isp went so dead it wouldn't even route to itself in the next city and remained that way for most of the afternoon.
Other than watching my boss turn new shades of red and screaming nothing much happend and life went on.
Case in point a few friends were ridding in the suzuki version of the Geo Metro and didn't have space in the car for the PS/2 so they put it on the roof and someone put his arm out the window to hold it.
The get most of the way home when the thing blows off the roof while the car is going 110Km/hour and bounces twice on the shoulder before going into the ditch.
They stop and pick it up and when they get home they plug it in.. still works.
Started self taught, got my first full time job and wsa left alone to run 30 servers all set by someone who quit rather than maintain the mess. I very quickly realised I knew nothing but learned fast.
Unix is a registerd trademark ad can only be used by software derived from the original AT&T code. Using it anywhere else tends to get threats of lawsuits.
"A 200 Meg / is pretty big. It would be better to simply mount/tmp seperately or in memory if it will be that busy. Also that way / cannot fill up with crap from/tmp"
I agree completely I wish he had done that.
"As for a 2Gb/var that is a large/var. If you really have that much logging you would have logging done to a seperate disk, ideally on a seperate channel to reduce I/O problems with the rest of the system."
I just remembered what else he used that space for(knew I was forgetting something): he also put the mysql dbs in there.
While I freely admit to having my own perfered style I also prefer not waste time fixing what's already working.
That's why when I arrive at a new jobsite I prefer not to make drastic changes for the first coupple of weeks while I get a feel for what is already there. By not reformatting everything right away I have a chance to learn how things were working, and will know what sucks and more importantly if they did anything I can learn from and adopt into my way of doing things.
Drastic changes for no reason(and even for a lot of good reasons) will not make my boss happy. Downtime is bad.
It's also been completely unavoidable in some cases. A mark of a good sysadmin is to know how his/her setup will handle growth. example: At one workplace I found a system that was with small / and/var partitions. After some time he ran out of space in/tmp so he moved it to/var/tmp and symlinked. He again ran out of space so he moved tmp to/usr/local/tmp and added a symlink from/var/tmp giving it a symlink to a symlink. Then from the looks of it/var filled up again.. so he created a/usr/local/var and created the required symlinks. So now I get the system and find that we now have a mostly unused partition and a tangle of symlinks. The sad thing is that he never learned from this. I again found a setup like that when taking over one of his systems at a later jobsite. The real laugh came at the second jobsite where he had moved to programming and refused to do a new project until I reformatted a machine he had setup in the past.
Thanks to people like that I tend to look good.. both jobsites went from weekly unplanned outages to rock solid.
If only that were true. Unfortunatly Unix admins have the same variance. I've had several jobs where at first much of my time was spent cleaning up after the incompetant who had the job before me.
The advantage of a Unix admin is that (s)he can make much more efficiant use of their time.
The "lets try it and see if it works" spammers don't last long. I've seen what happens first hand.
They try it (possibly over the objections of their tech staff) Get smacked down by the ISP and they either listen or they don't. If they don't they get their plug pulled.
So were already stuck with the hard core spammers and that's what bothers me. This plan as proposed would only serve to increase the spam being sent since it lets the other wannabes back into the market and I fail to see how that is of any advantage to the average consumer at all.
I have no idea how these people think spammers would volunteer to be regulated and pay more. Even given the prescribed system the current bulk mailing methods will still work and still be cheaper.
I seriously doubt most of these guys care at all about regulations or laws given the lame illegel or immoral crap I see flooding my inbox.
For that it does work and cleanly since on any decent 64 bit arch memory access is flat(yes I know there are sick exceptions). Since flat memory tends to agree with the way Linux apps expect memory to be addressed anything compiled for 64 bit should have no problem addressing all available memory.
The support was pretty much a byproduct of porting Linux to that arch.
I'm looking forward to the AA-64 as well since it fixes the 2 largest pains of the x86 world: The other problem being the lack of registers.
It's a crappy hack on Linux because it's a crappy hack on the archetecture itself. The limits of 32 bit addressing happens to be about 4 gig and to get more than that you get to use an EMS like thing that expands the address to 48 bits by paging things in and out of the 32 bit addressable space. By limiting each process to 4 gig they avoided having to pull off an even uglier hack that would have been required to export PAE to userspace.
Don't expect Linux to be any less ugly about something that works in a completely different way from how everything else does.
While good in the long term dumping x86 would be a huge pain for their customers who often have invested large amounts of money in applications on the x86 line. Dumping x86 would be a huge pain that would leave them open to AMD or VIA just walking in and filling existing customer demand with faster new x86 chips.
I also don't know where you get the impression that SUN doesn't do 64 bit well Sun's CPUs tend to preform very well although they are now having their bottom end cut out by x86.
What isn't doing well is the new Itanium.. The only reason it's even preforming close to par with competing CPUs is the huge die size the threw at it ensuring it will be a long time before it can ever be priced low enough for the desktop market. On top of that the Itanium demands the compilor and OS do things not required on competing platforms to accieve good performance.
Dumping x86 for itanium is just too dangerous even though a lot of people wish for the end of x86.
Unless you specially write your app to handle the oddities of handing 64GB (assuming the OS even allows you to) your limmited for 4GB per process since that's all you can address without resorting to the pentium's equivelant to EMS. The OS can hide the complexity and provide 64GB total but even then your stuck with 4GB per process.
Slashdot Mirror
Sure blame the victim.. is it our fault we happened to host something the kiddies thought they could get braging rights for taking down?
No the internet is *not* working well these days.. it's working well until you piss off some kiddy who has a lot of waiting drones. Or until some worm hits.
.. so be it. It *has* to be better than what we have now.
And the problem is not one of archetecture it's one of lazy admins who can't be bothered to upgrade in 4 months and of management who demand applications based on vendors with poor security records.
I am no longer happy to have to make the needed phone calls when my stuff is being flooded offline. I'm at the point where I just want someone to give the legions of the lazy and incompetant to go off and find some other line of work where they don't have to affect the rest of us. If that means govt enforced fines or licensing
Probably better odds elsewhere. After having actually seen what passes for security in the online gambling industry all you need really is a java decompilor.
Many of the online casino games tell the server whether it won or lost. And on one particularly funny case the game connected right to the SQL server at the casino.
The first place I worked did better with a flash/php combo but theve never bothered to upgrade apache or ssl since I quit.
It's quite sad really I'm supprised these places don't get ripped off more often.
Yeah well I too miss the days when a rooted server on someone else's network was not my problem. But welcome to today.
How exactly are network archetects supposed to design for 300 drones all sending traffic to one place? There is no amount of overcapacity that would handle that.
This is getting to less and less be the case. Keep in mind that the traffic caused by the slammer worm managed to disrupt 911 services.
.. what is your mom doing running servers? If there is no one to maintain her systems then there should be no outside accessable daemons at all.
Also
It's certainly not as bad as they are making out. It's only been a few months since one of worldcom's border routers went bad and took out several other isp's with it. My uplink isp went so dead it wouldn't even route to itself in the next city and remained that way for most of the afternoon.
Other than watching my boss turn new shades of red and screaming nothing much happend and life went on.
The problem is though that I got used to those machines.
I was discusted the first time I got a pentium(a clone) only to discover that by leaning my feet on the tower I had actually bent the case and frame.
Yes and I still have one.. unfortunatly it's on the other side of the country at my parent's house.
IBM used to make very solid hardware.
Case in point a few friends were ridding in the suzuki version of the Geo Metro and didn't have space in the car for the PS/2 so they put it on the roof and someone put his arm out the window to hold it.
The get most of the way home when the thing blows off the roof while the car is going 110Km/hour and bounces twice on the shoulder before going into the ditch.
They stop and pick it up and when they get home they plug it in.. still works.
Pity they don't make them that solidly anymore.
Rest of my rent is off here
Started self taught, got my first full time job and wsa left alone to run 30 servers all set by someone who quit rather than maintain the mess. I very quickly realised I knew nothing but learned fast.
So I'd say I learned by fire.
Unix is a registerd trademark ad can only be used by software derived from the original AT&T code. Using it anywhere else tends to get threats of lawsuits.
"A 200 Meg / is pretty big. It would be better to simply mount /tmp seperately or in memory if it will be that busy. Also that way / cannot fill up with crap from /tmp"
/var that is a large /var. If you really have that much logging you would have logging done to a seperate disk, ideally on a seperate channel to reduce I/O problems with the rest of the system."
I agree completely I wish he had done that.
"As for a 2Gb
I just remembered what else he used that space for(knew I was forgetting something): he also put the mysql dbs in there.
"Please do. I find it both interesting and informative."
That's moving off topic so I'll move that list here
To make sure there is enough space in each partition to handle the task that your going to throw at the server.
/tmp and /tmp is going to handle a lot of working files from the server.
/var if you have a highly loaded apache throwing logs at it.
You can't have a 200 meg / if it it's going to include
You can't have a 2 gb
Things like that. There were other issues but then I'd start ranting and go on for the next hour.
While I freely admit to having my own perfered style I also prefer not waste time fixing what's already working.
/var partitions. After some time he ran out of space in /tmp so he moved it to /var/tmp and symlinked. He again ran out of space so he moved tmp to /usr/local/tmp and added a symlink from /var/tmp giving it a symlink to a symlink. Then from the looks of it /var filled up again .. so he created a /usr/local/var and created the required symlinks. So now I get the system and find that we now have a mostly unused partition and a tangle of symlinks. The sad thing is that he never learned from this. I again found a setup like that when taking over one of his systems at a later jobsite. The real laugh came at the second jobsite where he had moved to programming and refused to do a new project until I reformatted a machine he had setup in the past.
That's why when I arrive at a new jobsite I prefer not to make drastic changes for the first coupple of weeks while I get a feel for what is already there. By not reformatting everything right away I have a chance to learn how things were working, and will know what sucks and more importantly if they did anything I can learn from and adopt into my way of doing things.
Drastic changes for no reason(and even for a lot of good reasons) will not make my boss happy. Downtime is bad.
It's also been completely unavoidable in some cases. A mark of a good sysadmin is to know how his/her setup will handle growth. example: At one workplace I found a system that was with small / and
Thanks to people like that I tend to look good.. both jobsites went from weekly unplanned outages to rock solid.
If only that were true. Unfortunatly Unix admins have the same variance. I've had several jobs where at first much of my time was spent cleaning up after the incompetant who had the job before me.
The advantage of a Unix admin is that (s)he can make much more efficiant use of their time.
By moving most of the work to the client instead of the server it would lower their back end costs as well.
The "lets try it and see if it works" spammers don't last long. I've seen what happens first hand.
They try it (possibly over the objections of their tech staff) Get smacked down by the ISP and they either listen or they don't. If they don't they get their plug pulled.
So were already stuck with the hard core spammers and that's what bothers me. This plan as proposed would only serve to increase the spam being sent since it lets the other wannabes back into the market and I fail to see how that is of any advantage to the average consumer at all.
I have no idea how these people think spammers would volunteer to be regulated and pay more. Even given the prescribed system the current bulk mailing methods will still work and still be cheaper.
I seriously doubt most of these guys care at all about regulations or laws given the lame illegel or immoral crap I see flooding my inbox.
For that it does work and cleanly since on any decent 64 bit arch memory access is flat(yes I know there are sick exceptions). Since flat memory tends to agree with the way Linux apps expect memory to be addressed anything compiled for 64 bit should have no problem addressing all available memory.
The support was pretty much a byproduct of porting Linux to that arch.
I'm looking forward to the AA-64 as well since it fixes the 2 largest pains of the x86 world: The other problem being the lack of registers.
It's a crappy hack on Linux because it's a crappy hack on the archetecture itself. The limits of 32 bit addressing happens to be about 4 gig and to get more than that you get to use an EMS like thing that expands the address to 48 bits by paging things in and out of the 32 bit addressable space. By limiting each process to 4 gig they avoided having to pull off an even uglier hack that would have been required to export PAE to userspace.
Don't expect Linux to be any less ugly about something that works in a completely different way from how everything else does.
While good in the long term dumping x86 would be a huge pain for their customers who often have invested large amounts of money in applications on the x86 line. Dumping x86 would be a huge pain that would leave them open to AMD or VIA just walking in and filling existing customer demand with faster new x86 chips.
I also don't know where you get the impression that SUN doesn't do 64 bit well Sun's CPUs tend to preform very well although they are now having their bottom end cut out by x86.
What isn't doing well is the new Itanium.. The only reason it's even preforming close to par with competing CPUs is the huge die size the threw at it ensuring it will be a long time before it can ever be priced low enough for the desktop market. On top of that the Itanium demands the compilor and OS do things not required on competing platforms to accieve good performance.
Dumping x86 for itanium is just too dangerous even though a lot of people wish for the end of x86.
Unless you specially write your app to handle the oddities of handing 64GB (assuming the OS even allows you to) your limmited for 4GB per process since that's all you can address without resorting to the pentium's equivelant to EMS. The OS can hide the complexity and provide 64GB total but even then your stuck with 4GB per process.
Did it here too only in C as an acidblood module.. end goal is a character sheet and battle moderator.
It's been fun.
I've seen some but the problem is the complete lack of door games for linux.
Most of the BBS software hasn't been updated in the past few years so tends to be for either DOS or windows.
I actually tried it with DOSemu but had a problem with connections remaining open if the player kills the telnet sesstion without logging out.