Shared hosting usually accomplishes this by terminating the SSL connection at a load balancer, then dispatching the request unencrypted (behind a firewall and in private IP space) to one of a series of webservers, and re-encrypting the response. The vhosts on the webserver are as you describe, but can serve thousands and thousands of hosts.
Practically, you do need a dedicated VIP (not necessarily server) to support non-wildcard SSL. Wildcard SSL is difficult for many hosting providers to mandate, as customers will want to use their own domainname to reinforce their branding.
wrt out-of-state drivers: I'm sure California will implement the scheme such that a base tax "penalty" will apply at the pump for drivers that don't have the GPS unit, and the GPS monitoring system would serve to "credit" the tax.
Serious problems in O'Reilly Editorship
on
Firefox In Print
·
· Score: 3, Insightful
O'Reilly has successfully transitioned from a geek publisher to yet another corporate sellout. Quality of content has really tanked, and even those few geek-oriented books that do get released are woefully thin volumes (W. Curtis Preston, whose fantastic O'Reilly Backup book should be considered the bible of backup and restore, can't write more than 200 pages on NAS and SAN? I think the topic's a bit broader than can be covered in such a thin tome).
Tim, if you're reading this, help restore O'Reilly to the kick-ass publisher of days of yore. Kill the Hacks books. Get rid of the Annoyances. Lose the Missing Manuals. Forget about the Notebooks. Concentrate on the Nutshells and the Essentials and the Animal Books (Pocket References are good, as well). Make them well-written, well-constructed, accurate, fun, and RELEVANT. Examples of excellence: Sendmail, DNS and BIND, Unix Backup and Recovery.
There are a bunch of self-righteous egotists who hang out here and contend that they just shouldn't have access to technology.
One significant social problem is indeed naive and therefore dangerous internet users. Just as it's generally unwise to give the keys to a nine-year-old, so too it's not reasonable to reduce barriers to the internet such that they decrease the quality of the internet in general. Right now, the Internet's not looking like the Information Super Highway so much as it looks like the Highway of Death out of Kuwait in Gulf War I.
It should be necessary to get a license (and therefore some training) before you're given an IP address and an email account.
They weren't looking for a file-sharing solution, or peer-to-peer system. They were looking for a distributed case management system, which is quite a different beast.
One problem I see is that companys demand any invention be turned over to themselves as terms of employment.
Decent patent reform would make this illegal - any invention created by an employee under the guise of his job would still be patented and owned by the employee, and the company who funded the invention would automatically get license rights to that patent over the term of the patent IF they pay the filing fee, but the employee could shop it around, develop it himself, etc - but NOT transfer ownership (destroy these scummy patent brokerage houses).
In my view, such reform would respect the original intention of patent laws: to reward and encourage innovation.
I know you're joking, but for the sake of a few sentences, let's pretend you're not. NO. If you want to slow down overall internet mail delivery, shut down hotmail.com, msn.com, yahoo.com, aol.com - and watch large legitimate mail gateways start to back up with thousands upon thousands of deferred messages.
Oh, I dunno - seems like they contribute quite a lot of value to the Linux kernel, and Satellite server plus kickstart make real differences when managing a few hundred machines. I disagree with a lot of their distribution policies (price definitely being one of them), but I don't mind funding Linux through RedHat (though I feel you're on mark about customers feeling extorted - it's overpriced.)
3 digits most certainly is something when you have a few hundred servers (and RedHat's support is a bit lukewarm - I've seen worse, but that's hardly high praise). I don't mind giving RedHat something for their efforts: Satellite is way cool, they continue to add a lot of value to the Linux kernel (check the kernel changelogs some time), and they deserve to be compensated - but 2 digits is a lot more palatable.
If you're comparing RISC Unix sales, yes, the top sellers are AIX, HP-UX, and Solaris, all of which are miniscule compared to OSes for Intel-type architectures (I base this on eye-ball observations in large colocation centers).
HP-UX will go away. AIX will go away. Solaris will go away. Linux will go away. Windows will go away. What order they start dropping off is up in the air, but probably the first three are pretty close to accurate.
And NFS had (and has) lots of other problems. You won't convince me that "NFS was a good design at the time"; I was there at the time, and I know it wasn't.
I didn't call NFS a good design, I called it a pragmatic solution - implying there are problems. Warts and all, it mostly works and is easy, and has usefulness. NFS succeeded because its usefulness in a wide variety of environments exceed its lack of security (I'm quite aware of those other problems you mention as well, mostly pertaining to locking, file size limitation, and multi-writer semantics [IIRC, last write wins])
No particular beef with your post - as you mention, NFS security isn't good. Put another way, it's as good as your network and host security, but relies on proper configuration of server and client. That said, NFS for being the pile of shit it is has survived through virtue of being widely supported and mostly compatible, robust in the face of server outages (as opposed to SMB's "oh, I guess you'll be losing the latest copy of your work now"), easily clusterable, scalable to several dozen busy clients, easy to manage, very well understood, and free.
No one in the know would hold the NFS security model up as one to emulate (you're of course absolutely correct about its security only being up to physical control of every machine on your network), but facts are, NFS continues to help people get the work they care about done. One can only hope that better solutions might finally be implemented to drive a worn-out protocol out of business, but it isn't there yet (I'd like it to be AFS, but on Linux, scalability doesn't seem to be on par with NFS).
NFS was a pragmatic solution designed by smart people in a time where talk, finger, telnet, RIP and rsh were all used freely. Now that networks are chock-a-block with assholes, NFS looks about as good as those other protocols - but people look a lot worse.
Um, I hate to tell you this, but while the US$ may not be "real" in the sense that it directly represents an actual commodity, there is no less trust involved in a gold-backed currency.
...which was surely the reasoning behind switching from the gold standard in the first place.
What incentive do printer manufacturers have to do this? It seems unlikely that the government would be able to pass legislation REQUIRING them to implement this technology, and surely implementing this technology adds cost to the printer.
Oh, I think it'd be pretty easy to the get any government to require measures that would protect the stability of their currency. Keep in mind that there's nothing backing the dollar other than user belief that that piece of paper actually has inherent worth (and by user, here I mean The World That Takes Dollars And Converts Them To $50 DVD Players And Hondas). Seriously bad things would happen to a lot of economies should the dollar freefall into worthlessness.
Remember, the dollar buys a soda because people believe it can.
I think you might be wrong on Dell not using AMD chips in the near future. The difference between performance between Intel and AMD offerings is very real, and their server line is being savaged in some markets by offerings from HP, IBM, and Sun.
Intel meanwhile seems to be in serious trouble - high clock-speed offerings canceled, Itanium in nowhere-ville, bottom end being munched by AMD's Duron. You could be right on Dell trying to twist some price advantage - if I was Intel presently, I wouldn't be very happy...
A very good article came out recently in Sysadmin documenting the issues of PC server performance - while lots has been done on processor architecture, this article talks about system architecture: http://www.samag.com/documents/s=9408/sam0411b/041 1b.htm
I went back to school for a NeXT (this was the bargain with the missus, get a degree -> get a NeXT).
The NeXT was QUITE interesting at the time - the 68030/40 was not a bad chip at the time, NeXT added interesting hardware like the Motorola 56000 DSP, Display Postscript was as you said very interesting, tons of custom silicon on the NeXT, interesting and for the time BLAZING fast 400dpi laser printer (there were no 600 dpi laser printers at the time). Again as you mention, the optical drives were really unique, and should have always been used as supplementary media rather than as boot media (this lesson got learned by the time of NeXT 040 and slabs). This machine tried hard to really bend what was possible.
You are very right about the APPs - nice bundling there. Underlying OS was pretty stable - we had a pair of NeXT slabs running for 1100 days at one point that shared/usr, and we'd completely forget about them - take the network down, yank DNS and YP servers out from under them - they just didn't care, they'd keep running.
NetInfo was way too obtuse to catch on, but a valiant effort at solving the NIS-is-crap problem, and ObjectiveC with InterfaceBuilder effectively created the RAD industry. NeXTStep was (and is, in current MacOS X implementation) quite cool, if not quite fast. To my thinking, Objective-C programming is an elegant solution to object-oriented programming (much nicer than C++), though certainly not perfect (run-time dynamic library resolution was a surprise to me, and Iwas disappointed that more invisible memory-handling features weren't provided).
to clarify: marketing execs are twisting that knob on the side of the Star Wars money machine, minting 10s and 20s, and are doing so by tired, rehashed and most importantly mediocre "properties" that demonstrate only too clearly their contempt for the very fans that are providing the honey. The Star Wars games that I've seen haven't been horrid - they've just been lacking new substance and long-term playability.
... the problem with the Star Wars franchise in a nutshell:
"there will be new Star Wars properties on TV"
The folks running this thing don't give one shit about the universe, it's not "shows", it's "properties" - that which may be bought and sold, rather than created, nurtured, or designed to inspire.
With the homogenization of radio ala clearchannel, and their annoying "demographics" and pay-for-play formats, the biggest problem I have when I DO hear the rare song that catches my ear is that I have no way of finding out what it is I just listened to.
iTunes and its ilk have made purchasing new music a bit more viable for me now - the previews are good (I'd buy half the tracks on every Cake album out there, and discard the rest) and I appreciate the "people who bought X also bought Y" references - it turns me onto some tunes and bands I hadn't heard of before.
Similarly, the in-store preview audio systems like RedDot in Barnes & Noble and Borders are pretty useful also.
Corporations need to get over the whole physical media thing, plus make radio a useful method of getting new music out there (why they play the same 50 songs all day long blows my mind - no request mechanisms, no way to throw out new music to expose potential big sellers - it's music by committee)..
quote: "storage capacity of 1-10 GB of data in under 1 cm^2 area with access times of under a millisecond and streaming bandwidths of over 50 Mbytes per second."
The research is about 5 years old. Because of constant seek times (the surface agitates in both x and y axes) and a kajillion heads, this is technology really designed to bridge EEPROM versus hard drive access times/throughput.
Think 50 Mbytes per second isn't any great shakes? Keep in mind that this is a chip less than a square centimeter in area, and start thinking of replacing RAID drives with these.
Humor all over there - from the baddies waving their butts at you and chortling, carrying bullseye targets around and holding them in front of their faces, peering out behind them every now and then, to hefting huge bombs around they can barely carry, to faking injury and limping off after a few shots.
Slashdot Mirror
You describe one possible scenario.
Shared hosting usually accomplishes this by terminating the SSL connection at a load balancer, then dispatching the request unencrypted (behind a firewall and in private IP space) to one of a series of webservers, and re-encrypting the response. The vhosts on the webserver are as you describe, but can serve thousands and thousands of hosts.
Practically, you do need a dedicated VIP (not necessarily server) to support non-wildcard SSL. Wildcard SSL is difficult for many hosting providers to mandate, as customers will want to use their own domainname to reinforce their branding.
Any of the above databases will suffer the same problems if write-cache on the disks used for storage is turned on.
wrt out-of-state drivers: I'm sure California will implement the scheme such that a base tax "penalty" will apply at the pump for drivers that don't have the GPS unit, and the GPS monitoring system would serve to "credit" the tax.
O'Reilly has successfully transitioned from a geek publisher to yet another corporate sellout. Quality of content has really tanked, and even those few geek-oriented books that do get released are woefully thin volumes (W. Curtis Preston, whose fantastic O'Reilly Backup book should be considered the bible of backup and restore, can't write more than 200 pages on NAS and SAN? I think the topic's a bit broader than can be covered in such a thin tome).
Tim, if you're reading this, help restore O'Reilly to the kick-ass publisher of days of yore. Kill the Hacks books. Get rid of the Annoyances. Lose the Missing Manuals. Forget about the Notebooks. Concentrate on the Nutshells and the Essentials and the Animal Books (Pocket References are good, as well). Make them well-written, well-constructed, accurate, fun, and RELEVANT. Examples of excellence: Sendmail, DNS and BIND, Unix Backup and Recovery.
One significant social problem is indeed naive and therefore dangerous internet users. Just as it's generally unwise to give the keys to a nine-year-old, so too it's not reasonable to reduce barriers to the internet such that they decrease the quality of the internet in general. Right now, the Internet's not looking like the Information Super Highway so much as it looks like the Highway of Death out of Kuwait in Gulf War I.
It should be necessary to get a license (and therefore some training) before you're given an IP address and an email account.
sloth jr
They weren't looking for a file-sharing solution, or peer-to-peer system. They were looking for a distributed case management system, which is quite a different beast.
What possible beneficial effect does false repudiation offer? More ways for people to be assholes...
One problem I see is that companys demand any invention be turned over to themselves as terms of employment.
Decent patent reform would make this illegal - any invention created by an employee under the guise of his job would still be patented and owned by the employee, and the company who funded the invention would automatically get license rights to that patent over the term of the patent IF they pay the filing fee, but the employee could shop it around, develop it himself, etc - but NOT transfer ownership (destroy these scummy patent brokerage houses).
In my view, such reform would respect the original intention of patent laws: to reward and encourage innovation.
I know you're joking, but for the sake of a few sentences, let's pretend you're not. NO. If you want to slow down overall internet mail delivery, shut down hotmail.com, msn.com, yahoo.com, aol.com - and watch large legitimate mail gateways start to back up with thousands upon thousands of deferred messages.
Oh, I dunno - seems like they contribute quite a lot of value to the Linux kernel, and Satellite server plus kickstart make real differences when managing a few hundred machines. I disagree with a lot of their distribution policies (price definitely being one of them), but I don't mind funding Linux through RedHat (though I feel you're on mark about customers feeling extorted - it's overpriced.)
3 digits most certainly is something when you have a few hundred servers (and RedHat's support is a bit lukewarm - I've seen worse, but that's hardly high praise). I don't mind giving RedHat something for their efforts: Satellite is way cool, they continue to add a lot of value to the Linux kernel (check the kernel changelogs some time), and they deserve to be compensated - but 2 digits is a lot more palatable.
sloth jr
If you're comparing RISC Unix sales, yes, the top sellers are AIX, HP-UX, and Solaris, all of which are miniscule compared to OSes for Intel-type architectures (I base this on eye-ball observations in large colocation centers).
HP-UX will go away. AIX will go away. Solaris will go away. Linux will go away. Windows will go away. What order they start dropping off is up in the air, but probably the first three are pretty close to accurate.
sloth jr
I didn't call NFS a good design, I called it a pragmatic solution - implying there are problems. Warts and all, it mostly works and is easy, and has usefulness. NFS succeeded because its usefulness in a wide variety of environments exceed its lack of security (I'm quite aware of those other problems you mention as well, mostly pertaining to locking, file size limitation, and multi-writer semantics [IIRC, last write wins])
No particular beef with your post - as you mention, NFS security isn't good. Put another way, it's as good as your network and host security, but relies on proper configuration of server and client. That said, NFS for being the pile of shit it is has survived through virtue of being widely supported and mostly compatible, robust in the face of server outages (as opposed to SMB's "oh, I guess you'll be losing the latest copy of your work now"), easily clusterable, scalable to several dozen busy clients, easy to manage, very well understood, and free.
No one in the know would hold the NFS security model up as one to emulate (you're of course absolutely correct about its security only being up to physical control of every machine on your network), but facts are, NFS continues to help people get the work they care about done. One can only hope that better solutions might finally be implemented to drive a worn-out protocol out of business, but it isn't there yet (I'd like it to be AFS, but on Linux, scalability doesn't seem to be on par with NFS).
NFS was a pragmatic solution designed by smart people in a time where talk, finger, telnet, RIP and rsh were all used freely. Now that networks are chock-a-block with assholes, NFS looks about as good as those other protocols - but people look a lot worse.
sloth jr
sloth jr
Oh, I think it'd be pretty easy to the get any government to require measures that would protect the stability of their currency. Keep in mind that there's nothing backing the dollar other than user belief that that piece of paper actually has inherent worth (and by user, here I mean The World That Takes Dollars And Converts Them To $50 DVD Players And Hondas). Seriously bad things would happen to a lot of economies should the dollar freefall into worthlessness.
Remember, the dollar buys a soda because people believe it can.
Wait a minute - this is a review posted on a MANUFACTURER'S site. GamePC sells PCs, including, surprise, a Pentium M gaming system.
I think you might be wrong on Dell not using AMD chips in the near future. The difference between performance between Intel and AMD offerings is very real, and their server line is being savaged in some markets by offerings from HP, IBM, and Sun.
1 1b.htm
Intel meanwhile seems to be in serious trouble - high clock-speed offerings canceled, Itanium in nowhere-ville, bottom end being munched by AMD's Duron. You could be right on Dell trying to twist some price advantage - if I was Intel presently, I wouldn't be very happy...
A very good article came out recently in Sysadmin documenting the issues of PC server performance - while lots has been done on processor architecture, this article talks about system architecture:
http://www.samag.com/documents/s=9408/sam0411b/04
sloth_jr
I went back to school for a NeXT (this was the bargain with the missus, get a degree -> get a NeXT).
/usr, and we'd completely forget about them - take the network down, yank DNS and YP servers out from under them - they just didn't care, they'd keep running.
The NeXT was QUITE interesting at the time - the 68030/40 was not a bad chip at the time, NeXT added interesting hardware like the Motorola 56000 DSP, Display Postscript was as you said very interesting, tons of custom silicon on the NeXT, interesting and for the time BLAZING fast 400dpi laser printer (there were no 600 dpi laser printers at the time). Again as you mention, the optical drives were really unique, and should have always been used as supplementary media rather than as boot media (this lesson got learned by the time of NeXT 040 and slabs). This machine tried hard to really bend what was possible.
You are very right about the APPs - nice bundling there. Underlying OS was pretty stable - we had a pair of NeXT slabs running for 1100 days at one point that shared
NetInfo was way too obtuse to catch on, but a valiant effort at solving the NIS-is-crap problem, and ObjectiveC with InterfaceBuilder effectively created the RAD industry. NeXTStep was (and is, in current MacOS X implementation) quite cool, if not quite fast. To my thinking, Objective-C programming is an elegant solution to object-oriented programming (much nicer than C++), though certainly not perfect (run-time dynamic library resolution was a surprise to me, and Iwas disappointed that more invisible memory-handling features weren't provided).
sloth jr
to clarify: marketing execs are twisting that knob on the side of the Star Wars money machine, minting 10s and 20s, and are doing so by tired, rehashed and most importantly mediocre "properties" that demonstrate only too clearly their contempt for the very fans that are providing the honey. The Star Wars games that I've seen haven't been horrid - they've just been lacking new substance and long-term playability.
"there will be new Star Wars properties on TV"
The folks running this thing don't give one shit about the universe, it's not "shows", it's "properties" - that which may be bought and sold, rather than created, nurtured, or designed to inspire.
Big gaping pit of money-grubbing marketers.
With the homogenization of radio ala clearchannel, and their annoying "demographics" and pay-for-play formats, the biggest problem I have when I DO hear the rare song that catches my ear is that I have no way of finding out what it is I just listened to.
iTunes and its ilk have made purchasing new music a bit more viable for me now - the previews are good (I'd buy half the tracks on every Cake album out there, and discard the rest) and I appreciate the "people who bought X also bought Y" references - it turns me onto some tunes and bands I hadn't heard of before.
Similarly, the in-store preview audio systems like RedDot in Barnes & Noble and Borders are pretty useful also.
Corporations need to get over the whole physical media thing, plus make radio a useful method of getting new music out there (why they play the same 50 songs all day long blows my mind - no request mechanisms, no way to throw out new music to expose potential big sellers - it's music by committee)..
sloth jr
For a background on the technology, check out:
http://yogi.pdl.cmu.edu/research/MEMS/
quote: "storage capacity of 1-10 GB of data in under 1 cm^2 area with access times of under a millisecond and streaming bandwidths of over 50 Mbytes per second."
The research is about 5 years old. Because of constant seek times (the surface agitates in both x and y axes) and a kajillion heads, this is technology really designed to bridge EEPROM versus hard drive access times/throughput.
Think 50 Mbytes per second isn't any great shakes? Keep in mind that this is a chip less than a square centimeter in area, and start thinking of replacing RAID drives with these.
sloth jr
Humor all over there - from the baddies waving their butts at you and chortling, carrying bullseye targets around and holding them in front of their faces, peering out behind them every now and then, to hefting huge bombs around they can barely carry, to faking injury and limping off after a few shots.
http://www.msi.com.tw/program/products/slim_pc/slm /pro_slm_detail.php?UID=546
I saw this at Fry's recently - was pretty impressed, includes motherboard, Linux support unknown.