In that case, ANY up-n-running system with Internet access would help you out. You'd need to be able to write compatible media, of course, and using the same HDD is obviously the easiest.
I got my first Linux PC running using a Mac and a questionable copy of DOS 5 that came with the surplus PC. (It had all of COMMAND.COM and the hidden system files.) This was before OS X, so the Mac didn't have a way of writing the boot disk image to floppy; but it would compress _just_ enough to fit on a DOS floppy, so I used my one (1) 1.4 MB diskette to get a copy of unzip, the boot disk image, and the DOS program for writing it to diskette on the PC.
Had that machine been able to boot from CD, no problem. I had a nice, bootable install CD. It was a crappy version of TurboLinux from a magazine cover, but it was enough to bootstrap a better distro without having to mess with DOS again.
The early Macs were mostly a CPU and the Toolbox ROM. They didn't have DMA, or complex on-board I/O; just a little SCSI controller, the desktop bus, and floppy. The video was just a bitmap in RAM, no line drawing, no blitter, nothing.
As a result, the Mac-on-Amiga emulators worked really, really well. One of them, I can't recall if it was the AMax, you plugged the Toolbox ROM chip into a dongle that then plugged into your parallel port.
I believe they actually could go a bit faster than the Macs of the day by using the Amiga's GPU.
The big killer, though, was the floppy drive. Because the Mac 800K floppy was actually a CLV drive, rather than CAV like everyone else, there was only about 30% of the diskette that could be used by both the Amiga and Mac floppy drives. So you needed to use a special disk format to move between a Real Mac and the Amiga.
...except that if Apple had earned themselvs a reputation for making each OS upgrade slower on current machines, they'd sell a whole lot fewer OS upgrades.
It would probably actually hurt them in lost hardware sales too, because a lot of people would just stick with a working configuration--and not update either the OS or the hardware. Skipping over issues around major design changeovers, like 604E to G3, Apple's OS upgrades let you keep an aging machine going along fairly comfortably--until YOU are ready to upgrade the hardware.
People like it when you let them make decisions in their own time.
It is really looking like the X86 cpu is reaching the end of it's life.
While I'm not a fan of the X86 architecture in general, or any of the chips in particular, it is important to keep in mind that what modern X86es have with earlier X86 chips is mainly the instruction stream.
AMD has shown how you can add new registers to an X86 chip while preserving execution compatibility for classic IA32 code. They also added 64-bit registers and instructions while preserving the 32-bit environment (much like SPARC, POWER and PowerPC did their 64-bit versions).
So, is it all that much of a stretch to imagine a mode flag that can be set by supervisor code that drops the IA32 instruction translator out of the pipeline, and starts pulling lower-level instructions for a particular process? All the other ideas are already there in AMD64: 32-bit classic, 32-bit updated with new registers and opcodes, and 64-bit, all timesliced onto the same CPU.
So, while I really don't care for the X86 family, I think it is far from dead.
And maybe removing the CISC decoder isn't that important anyway. Keep in mind the Xeons cache the decoded instruction for a given address, not the raw IA32 opcodes. So when you have an I-cache hit, you can skip 2-3 pipeline stages.
But I do think it would be amazing to see what the brainpower involved in keeping X86 alive could do if they started from scratch. As long as they weren't allowed to think of anything like the Itanium.
Yup, and the RS/6000 7015-900 series had a drawer full of batteries. They were basically the same rack design as the CISC AS/400.
And I see they still do, at least on the p5 590-class machines.
Smaller, single-drawer servers do not generally have integrated UPSes because it makes much more sense to have a UPS drawer in the rack to run the entire rack frame.
He got a new one from Business Depot oops I mean Staples.
Actually, it's a really good stapler, I've never used such a nice stapler. I can really understand his obsession with the Swingline stapler. And I got it just for... for... well, I had to. (And it even said, "The Perfect Complement to any Office Space" on the collectors edition packaging.)
And if they move my desk again, I'm going to burn the building down. I mean it.
Folks will get a really nasty surprise when your account is deleted,
Had a realization about unintentionally creating a situation like that at my previous job.
All the department's partitions on the AFS and DFS servers were charged to my account--they had no way of assigning space to a group. It was 4:30 PM before a long weekend. Very few people were left in IT.
I suddenly realized what would happen to all the batch jobs when everything belonging to my account was locked out.
My manager was able to find someone in IT who could suspend the automatic lockout until they could reassign all the filesystem resources...
Depends on how dodgy your hardware controllers are, I suppose. And not everyone wants "cool", some of us just want "works reliably".
I've run a fleet of AIX systems which had the equivelent of software RAID 1 (mirroring) and later added RAID 0 (striping) in the OS since at least the 3.1 days.
It is very, very, very convenient to be able to pull bog-standard drives into logical volumes with bog-standard controllers. Especially if management won't approve $5000 expenses for a $200,000 computer. (I'm not kidding--took me 4 years and hundreds of lost work-hours to get approval to buy UPSes for half a million dollars worth of computer. This was 10 years back, divide all the prices by 10, and multiply the performance by 20.)
And RAID-in-hardware really only matters with the parity calculated versions; if you just want cheap and quick redundancy, mirror your data onto two disks--it's not like that's expensive these days.
Another advantage of simple mirroring is access is very fast--there's no parity calculation required. (Though truly paranoid modes will read from both disks and compare the results; but in two-way mirroring, the only thing you really have to rely on is the I/O error indicators from the drive transport and checksumming.)
Blackdown is open source? Really? Wow, I can't find a single link to the source on their website. And Kaffe is really not up to snuff for arbitrary code.
We've got a source-available proprietary 3rd party library in our code. It's a C++ program. No big deal.
It's just their equivelent of./configure is written in Java. So, I've spent a big chunk of my time on the Linux on PowerPC port (to an RPM-based distribution, of course--Red Hat Enterprise Linux) in hunting down a JVM that is compatible with both the stupid configure program and the Linux distro in use.
So, even without being a GNU hardliner, the proprietary nature of Java is extremely frustrating--especially when the product you're working with doesn't even use Java once you get the damn thing built.
GPL is about freedom for the end-user--the one person who does NOT need to be bound by the GPL. (If they distribute the code, they are no longer (only) an end-user.) The GPL is there to make sure that one person gets a copy of the source needed to make the binary.
BSD is about freedom for the developer.
They've got different goals. The GPL was, essentially, written in response to buggy proprietary software that the supplier would not fix.
Yes, that's right. The PC, PC XT and PC AT keyboard ports and cassette port were on the mainboard, along with the "beep speaker". The cassette port might not have been present on all XT and AT machines. I'm not sure about the original PC, but the XT and AT both had floppy support on the mainboard too.
The PC AT introduced something really weird: they had a strange, stripped-down hard disk controller on the main board itself, that required drives with embedded control circuitry to function, much like SASI, SCSI or ESDI, and quite unlike the ST-506 or 412 setup that was more common.
It's legacy lives on, as the AT Attachment interface, or Integrated Drive Electronics. (And you can run SCSI commands over it, with the AT Attachment Packet Interface.)
In PS/2 land, the keyboard and mouse ports were always built in. Some models had integrated AT disk, some used ESDI controllers off the Microchannel bus, some used SCSI. I don't know if there was a MCA to IDE bus adapter.
And if anyone's still reading, keep the on-board network, USB and IDE interfaces. Use your video board; if you've got a good one, the integrated stuff doesn't come close. (And on my XP box, the integrated video gets a lot of weird distortion lines when playing DVD; I think the RAMDACs are having trouble getting to the main system RAM fast enough.)
Try the onboard sound, but be prepared to keep your existing card.
Other things, give them to friends with older machines. That's what I do.
That would be funnier if I hadn't had people actually try to communicate with me by typing like that. In e-mail or IM from a Real Keyboard(tm).
And not teenagers, either; late 20s.
It may be elitist of me, but after trying a few times, I decided that the cultural gap is just too big--I can't be friends with someone who doesn't even spell. I can barely get along with people who don't use capitals properly.
But, there are a lot of people out there, and I don't have to be friends with all of them.
Windows does it one way, Mac OS does it the other, and users on both platforms are used to that.
And now we're getting apps as part of OS X which are working the Windows way, which is really, really, really annoying.
It's one thing to not have something happen when you expect it to.
It's much, much worse to have something happen when you DON'T expect it to. Especially if you were trying to bring the app to the foreground so you could see what was selected so that you decide what to do next.
Near as I can tell, it's the stupid "metallic" apps that use the app-gets-focus-click, but I'm not sure. It's few enough apps that I haven't figured out the common thread yet, but it sure is annoying to have a few apps Work Different.
(Of course, X11 gets really confusing on OS X, so I just try to click somewhere safe with X11 windows.)
At least since AIX 3. I have had the fortune to never have administered an AIX 2 (RT/PC) or AIX 1 (PS/2) system, so I don't know about the previous incarnations. (AIX 4 was the first release of AIX that ran on the same hardware as its predecessor; though they did add PREP-compatible PowerPC machine support.)
The confusing part about AIX is that you wind up with some daemons started by srcmstr, some started as "respawn" jobs by init, some started by inetd, and some by good ol'/etc/rc scripts.
Apple seems to be fixing that, so at least you need to only know one way _per operating system_.
(Similarly, Windows seems to use "NET START" for everything that resembles a UNIX daemon, even if it has nothing to do with the 'net--net start "RunAs Service"; net start "iPod Service".)
I don't see the quality issues you're talking about.
Try the G3 dual USB iBooks, entire product family.
Granted, Apple is great at repairing the "Logic Board" issue free, but it is still annoying to be without your lapdog for a week or two every 3-4 months.
One of the main problems I have found at IBM--especially as a former employee but also as a customer now--is they will often make something complicated for the sake of making it complicated.
Apple's success has been in making it simple.
IBM has some very brilliant researchers, and some excellent programmers... and some from the other end of the scale, too. They will also often get bogged down in internal politics.
And, unlike Apple, IBM will do its best to give you several choices for everything. (Unfortunately, often none of the ones you want.) This helps their managers avoid making a qualitative decision--while Apple is more than happy to say, "You do it this way" and then make that way as slick and elegant as possible.
It got very frustrating working at IBM because of all this--we actually shipped a C++ compiler that didn't work with make! (I had a little Perl script which would preserve the incremental feature of the new compiler, and hook up to make, and then allow most programs to compile. We weren't allowed to ship it, because it didn't solve 100% of potential cases--so we had 100% annoyed customers, when it needed to be only 10% annoyed.)
That's because the ISO 9660 CD format only supports 8.3 filenames.
ISO 9660 supports 31 character monocase filenames, including the period. The period is mandatory.
However, many, many ISO 9660 drivers were written with the DOS 8.3 naming convention in mind, and will do everything from wedge your kernel to reboot your machine if you feed it a fully-conformant ISO 9660 disc.
Feel "responsive", maybe. When you're using a system what matters is quick feedback to your actions. So if it's still busy starting up services, and can't context-switch the GUI tasks in fast enough, it will feel unresponsive--and people will perceive that as slow. Since that same mass-market OS doesn't need to be rebooted very often, especially compared to its predecessors, waiting a minute or two before logging in gives you a more responsive system.
This is why the network guys are always talking about latency (time to start) vs. bandwidth (how much per unit time)--and the memory and CPU guys are too, these days--memory bandwidth is pretty good, but latency is a killer.
And, as a long-time iMac owner (iMac DV 400, the first iMac model with FireWire and VGA out), I can say it's great as your only computer.
But they suck bricks when you buy your second machine and want to hook them in to a KVM switch.
I'd be all over a G5 Mini based on an iMac G5 without the screen. So, until I can afford a really spiffy G5 tower that I'm not really sure I need anyway, I'll just get the current Mini. Real Soon Now.
(I have to wait a few weeks anyway, so let's see if they really do bump the base RAM to something sensible.)
You're most likely to run into problems with 127/8 addresses outside of 127.0.0/24. That is, some systems will have 127.0.0/24 as the loopback, instead of the whole class A range. Try pinging 127.1.1.1. (Seems OK on Linux and AIX.)
Check the netmask on your loopback adapter (ifconfig lo or ifconfig lo0, typically), see if it is 255.0.0.0; also check the routing table (netstat -nr) to see if there's a route for 127/8 going to 127.0.0.1. (Linux doesn't seem to bother, OS X and AIX both do.)
HP-UX seems to get it really wrong--they have a network route for 127/8, and the packets don't go through. OS X gets it wrong this way too.
Solaris gets it wrong but in a less-bad way--they don't have a network route, and the packets don't go through. But at least there isn't a route to say how to do it.
You used to see a lot of UNIX network config programs put "127.0.0.1 localhost full.dns.name" in/etc/hosts to do exactly that.
I always find it causes more trouble than it's worth to have an/etc/hosts entry conflict with my DNS config, so I take it back out.
More useful is putting "st.at.ic.ip full.dns.name" in/etc/hosts, if you've got a static IP anyway. (I'm pretty sure AIX does this for you, but it's been a while.) I suppose the DHCP client could take care of that for you, too.
Well, spammers cannot spam without either collusion or incompetence of ISP admins. They've got to reach the net somehow.
And yes, yanking an account on one or two spam complaints is wrong--and taking any action on a complaint which doesn't include enough information to verify that it is one of your customers is just stupid.
But how many spam runs are so small as to produce only one valid complaint? How many runs, especially for smaller ISPs, won't hammer the outbound SMTP relay into the ground?
Granted, use of "zombie relay" PCs means the problems aren't going to be where an ISP admin can see them. But that's what the "looks like a broadband" blocklist is for--and it is VERY effective. (It just means people on static DSL or Cable IPs who wish to use their own mail servers have to relay through their ISP's official server. I can live with that--inbound goes right to my machine, outbound goes to a reputable relay, so it doesn't come directly from a DSL node.)
But leaving spammers connected to the net, either as an upstream IP provider or as an actual ISP, does contribute to the spam problem.
And buying stuff from them (and from telemarketers who can't read a do-not-call list and door-to-door salesmen who don't understand "NO SOLICITING" on your front door) contributes to the problem even more. I welcome any suggestions on how to deal with people who actually buy from spammers.
Slashdot Mirror
I got my first Linux PC running using a Mac and a questionable copy of DOS 5 that came with the surplus PC. (It had all of COMMAND.COM and the hidden system files.) This was before OS X, so the Mac didn't have a way of writing the boot disk image to floppy; but it would compress _just_ enough to fit on a DOS floppy, so I used my one (1) 1.4 MB diskette to get a copy of unzip, the boot disk image, and the DOS program for writing it to diskette on the PC.
Had that machine been able to boot from CD, no problem. I had a nice, bootable install CD. It was a crappy version of TurboLinux from a magazine cover, but it was enough to bootstrap a better distro without having to mess with DOS again.
Not to deny the probability that "password harvesting fishing" is a backformation.
As a result, the Mac-on-Amiga emulators worked really, really well. One of them, I can't recall if it was the AMax, you plugged the Toolbox ROM chip into a dongle that then plugged into your parallel port.
I believe they actually could go a bit faster than the Macs of the day by using the Amiga's GPU.
The big killer, though, was the floppy drive. Because the Mac 800K floppy was actually a CLV drive, rather than CAV like everyone else, there was only about 30% of the diskette that could be used by both the Amiga and Mac floppy drives. So you needed to use a special disk format to move between a Real Mac and the Amiga.
387 cycles. I wonder why it doesn't seem to run as long on battery as it used to, it's only 2 1/2 years old.... (Max useful life is what, around 500?)
At least there's been more cycles on my battery than replacements of the iLogic iBoard.
It would probably actually hurt them in lost hardware sales too, because a lot of people would just stick with a working configuration--and not update either the OS or the hardware. Skipping over issues around major design changeovers, like 604E to G3, Apple's OS upgrades let you keep an aging machine going along fairly comfortably--until YOU are ready to upgrade the hardware.
People like it when you let them make decisions in their own time.
Just a question... when did IBM stop being the 1st largest software company in the world?
While I'm not a fan of the X86 architecture in general, or any of the chips in particular, it is important to keep in mind that what modern X86es have with earlier X86 chips is mainly the instruction stream.
AMD has shown how you can add new registers to an X86 chip while preserving execution compatibility for classic IA32 code. They also added 64-bit registers and instructions while preserving the 32-bit environment (much like SPARC, POWER and PowerPC did their 64-bit versions).
So, is it all that much of a stretch to imagine a mode flag that can be set by supervisor code that drops the IA32 instruction translator out of the pipeline, and starts pulling lower-level instructions for a particular process? All the other ideas are already there in AMD64: 32-bit classic, 32-bit updated with new registers and opcodes, and 64-bit, all timesliced onto the same CPU.
So, while I really don't care for the X86 family, I think it is far from dead.
And maybe removing the CISC decoder isn't that important anyway. Keep in mind the Xeons cache the decoded instruction for a given address, not the raw IA32 opcodes. So when you have an I-cache hit, you can skip 2-3 pipeline stages.
But I do think it would be amazing to see what the brainpower involved in keeping X86 alive could do if they started from scratch. As long as they weren't allowed to think of anything like the Itanium.
Yup, and the RS/6000 7015-900 series had a drawer full of batteries. They were basically the same rack design as the CISC AS/400.
And I see they still do, at least on the p5 590-class machines.
Smaller, single-drawer servers do not generally have integrated UPSes because it makes much more sense to have a UPS drawer in the rack to run the entire rack frame.
Actually, it's a really good stapler, I've never used such a nice stapler. I can really understand his obsession with the Swingline stapler. And I got it just for... for... well, I had to. (And it even said, "The Perfect Complement to any Office Space" on the collectors edition packaging.)
And if they move my desk again, I'm going to burn the building down. I mean it.
Had a realization about unintentionally creating a situation like that at my previous job.
All the department's partitions on the AFS and DFS servers were charged to my account--they had no way of assigning space to a group. It was 4:30 PM before a long weekend. Very few people were left in IT.
I suddenly realized what would happen to all the batch jobs when everything belonging to my account was locked out.
My manager was able to find someone in IT who could suspend the automatic lockout until they could reassign all the filesystem resources...
Depends on how dodgy your hardware controllers are, I suppose. And not everyone wants "cool", some of us just want "works reliably".
I've run a fleet of AIX systems which had the equivelent of software RAID 1 (mirroring) and later added RAID 0 (striping) in the OS since at least the 3.1 days.
It is very, very, very convenient to be able to pull bog-standard drives into logical volumes with bog-standard controllers. Especially if management won't approve $5000 expenses for a $200,000 computer. (I'm not kidding--took me 4 years and hundreds of lost work-hours to get approval to buy UPSes for half a million dollars worth of computer. This was 10 years back, divide all the prices by 10, and multiply the performance by 20.)
And RAID-in-hardware really only matters with the parity calculated versions; if you just want cheap and quick redundancy, mirror your data onto two disks--it's not like that's expensive these days.
Another advantage of simple mirroring is access is very fast--there's no parity calculation required. (Though truly paranoid modes will read from both disks and compare the results; but in two-way mirroring, the only thing you really have to rely on is the I/O error indicators from the drive transport and checksumming.)
Blackdown is open source? Really? Wow, I can't find a single link to the source on their website. And Kaffe is really not up to snuff for arbitrary code.
We've got a source-available proprietary 3rd party library in our code. It's a C++ program. No big deal.
It's just their equivelent of ./configure is written in Java. So, I've spent a big chunk of my time on the Linux on PowerPC port (to an RPM-based distribution, of course--Red Hat Enterprise Linux) in hunting down a JVM that is compatible with both the stupid configure program and the Linux distro in use.
So, even without being a GNU hardliner, the proprietary nature of Java is extremely frustrating--especially when the product you're working with doesn't even use Java once you get the damn thing built.
GPL is about freedom for the end-user--the one person who does NOT need to be bound by the GPL. (If they distribute the code, they are no longer (only) an end-user.) The GPL is there to make sure that one person gets a copy of the source needed to make the binary.
BSD is about freedom for the developer.
They've got different goals. The GPL was, essentially, written in response to buggy proprietary software that the supplier would not fix.
The PC AT introduced something really weird: they had a strange, stripped-down hard disk controller on the main board itself, that required drives with embedded control circuitry to function, much like SASI, SCSI or ESDI, and quite unlike the ST-506 or 412 setup that was more common.
It's legacy lives on, as the AT Attachment interface, or Integrated Drive Electronics. (And you can run SCSI commands over it, with the AT Attachment Packet Interface.)
In PS/2 land, the keyboard and mouse ports were always built in. Some models had integrated AT disk, some used ESDI controllers off the Microchannel bus, some used SCSI. I don't know if there was a MCA to IDE bus adapter.
And if anyone's still reading, keep the on-board network, USB and IDE interfaces. Use your video board; if you've got a good one, the integrated stuff doesn't come close. (And on my XP box, the integrated video gets a lot of weird distortion lines when playing DVD; I think the RAMDACs are having trouble getting to the main system RAM fast enough.)
Try the onboard sound, but be prepared to keep your existing card.
Other things, give them to friends with older machines. That's what I do.
And not teenagers, either; late 20s.
It may be elitist of me, but after trying a few times, I decided that the cultural gap is just too big--I can't be friends with someone who doesn't even spell. I can barely get along with people who don't use capitals properly.
But, there are a lot of people out there, and I don't have to be friends with all of them.
And now we're getting apps as part of OS X which are working the Windows way, which is really, really, really annoying.
It's one thing to not have something happen when you expect it to.
It's much, much worse to have something happen when you DON'T expect it to. Especially if you were trying to bring the app to the foreground so you could see what was selected so that you decide what to do next.
Near as I can tell, it's the stupid "metallic" apps that use the app-gets-focus-click, but I'm not sure. It's few enough apps that I haven't figured out the common thread yet, but it sure is annoying to have a few apps Work Different.
(Of course, X11 gets really confusing on OS X, so I just try to click somewhere safe with X11 windows.)
The confusing part about AIX is that you wind up with some daemons started by srcmstr, some started as "respawn" jobs by init, some started by inetd, and some by good ol' /etc/rc scripts.
Apple seems to be fixing that, so at least you need to only know one way _per operating system_.
(Similarly, Windows seems to use "NET START" for everything that resembles a UNIX daemon, even if it has nothing to do with the 'net--net start "RunAs Service"; net start "iPod Service".)
Try the G3 dual USB iBooks, entire product family.
Granted, Apple is great at repairing the "Logic Board" issue free, but it is still annoying to be without your lapdog for a week or two every 3-4 months.
Apple's success has been in making it simple.
IBM has some very brilliant researchers, and some excellent programmers... and some from the other end of the scale, too. They will also often get bogged down in internal politics.
And, unlike Apple, IBM will do its best to give you several choices for everything. (Unfortunately, often none of the ones you want.) This helps their managers avoid making a qualitative decision--while Apple is more than happy to say, "You do it this way" and then make that way as slick and elegant as possible.
It got very frustrating working at IBM because of all this--we actually shipped a C++ compiler that didn't work with make! (I had a little Perl script which would preserve the incremental feature of the new compiler, and hook up to make, and then allow most programs to compile. We weren't allowed to ship it, because it didn't solve 100% of potential cases--so we had 100% annoyed customers, when it needed to be only 10% annoyed.)
ISO 9660 supports 31 character monocase filenames, including the period. The period is mandatory.
However, many, many ISO 9660 drivers were written with the DOS 8.3 naming convention in mind, and will do everything from wedge your kernel to reboot your machine if you feed it a fully-conformant ISO 9660 disc.
This is why the network guys are always talking about latency (time to start) vs. bandwidth (how much per unit time)--and the memory and CPU guys are too, these days--memory bandwidth is pretty good, but latency is a killer.
But they suck bricks when you buy your second machine and want to hook them in to a KVM switch.
I'd be all over a G5 Mini based on an iMac G5 without the screen. So, until I can afford a really spiffy G5 tower that I'm not really sure I need anyway, I'll just get the current Mini. Real Soon Now.
(I have to wait a few weeks anyway, so let's see if they really do bump the base RAM to something sensible.)
Check the netmask on your loopback adapter (ifconfig lo or ifconfig lo0, typically), see if it is 255.0.0.0; also check the routing table (netstat -nr) to see if there's a route for 127/8 going to 127.0.0.1. (Linux doesn't seem to bother, OS X and AIX both do.)
HP-UX seems to get it really wrong--they have a network route for 127/8, and the packets don't go through. OS X gets it wrong this way too.
Solaris gets it wrong but in a less-bad way--they don't have a network route, and the packets don't go through. But at least there isn't a route to say how to do it.
I always find it causes more trouble than it's worth to have an /etc/hosts entry conflict with my DNS config, so I take it back out.
More useful is putting "st.at.ic.ip full.dns.name" in /etc/hosts, if you've got a static IP anyway. (I'm pretty sure AIX does this for you, but it's been a while.) I suppose the DHCP client could take care of that for you, too.
Well, spammers cannot spam without either collusion or incompetence of ISP admins. They've got to reach the net somehow.
And yes, yanking an account on one or two spam complaints is wrong--and taking any action on a complaint which doesn't include enough information to verify that it is one of your customers is just stupid.
But how many spam runs are so small as to produce only one valid complaint? How many runs, especially for smaller ISPs, won't hammer the outbound SMTP relay into the ground?
Granted, use of "zombie relay" PCs means the problems aren't going to be where an ISP admin can see them. But that's what the "looks like a broadband" blocklist is for--and it is VERY effective. (It just means people on static DSL or Cable IPs who wish to use their own mail servers have to relay through their ISP's official server. I can live with that--inbound goes right to my machine, outbound goes to a reputable relay, so it doesn't come directly from a DSL node.)
But leaving spammers connected to the net, either as an upstream IP provider or as an actual ISP, does contribute to the spam problem.
And buying stuff from them (and from telemarketers who can't read a do-not-call list and door-to-door salesmen who don't understand "NO SOLICITING" on your front door) contributes to the problem even more. I welcome any suggestions on how to deal with people who actually buy from spammers.