Of the things you listed, Oracle 9i is available on OS X; I'll give you Coldfusion; and Veritas isn't a network service, it's a freaking filesystem. I might as well point out that there's not a Linux box out there that can integrate into a netinfo domain -- while technically true, it doesn't matter in even the tiniest bit for network service provision.
In this case, hashing isn't going to discover anything at all interesting, with or without regular expressions. You could tokenize the source and build a parse tree (big fun with C! *pukes*), or taking a page from the information theory folks, use gzip, and, picking an arbitrary threshold of similarity, visually inspect the suspect chunks.
All three of which are wildly deserving of remakes.
'jfb
Re:The article misses a few things as well.
on
Sun's Last Stand
·
· Score: 1
Sun's chips are impressive but Intel's economy of scale kills them in price and comes close enough in performance that x86 is the smart decision for 99% of people today.
Good stuff, but one quibble: Sun's chips haven't been impressive, relative to their competition, since 1994. Everybody and their dog walks all over SPARC these days, and Intel is actually the absolute fastest, from P4 for 32bits to Itanic 2 for 64.
The only reason to spend the money on Solaris is if you're stuck with some tedious garbage like SAP. As DEC could tell you, that's not a good way to go about finding new customers. I give Sun two years.
You're better off with TERM set to dtterm, if'n your application supports it. Just in case you thought progress in computing had been made in the last 25 years, along comes a happy reminder like terminal handling to show you the error of your ways.
Terminal.app also emulates DEC's tres-cool dtterm. Just set your TERM variable to dtterm. It's a superset of xterm-color functionality, as near as I can tell. As it happens, you can make Terminal.app do all kinds of wacky things with escapes. For instance:
alias dock='echo -n "[2t"' alias lower='echo -n "[6t"' alias raise 'echo -n "[5t"'
PowerQuiccIII/MPC8560 supports not only DDR memory, but two Ghz ethernet controllers and a RapidIO controller.
... and no Altivec. It's not a G4 (MPC75xx family), it's a different class of processor entirely. That it implements Book E is of no consequence -- and besides, even with it's (single-channel) DDR, it's still slower and offers still less memory bandwidth than the 970.
The interesting thing was that raw number-crunching power wasn't always the most important thing -- many times it is bandwith...
... which points out in stark and amusing relief the differing worlds of desktop and embedded usage. The reason that Apple is so hot for the 970 is that it offers dual channel DDR memory -- quite a change from the 133mhz SDR that the G4 is limited to.
Well, you're half right. The POWER4 is designed for reliability. However, at 1.7ghz, it handily outperforms the projected numbers for a 970, particularly and unsurprisingly in floating point performance.
Don't get me wrong: as soon as a Mac with this baby in it is available, I'm upgrading, but let's call a spade a spade. The 970 looks to be decently faster than what we currently have in raw processing power, but with a radical, "holy cow where're my pants" faster memory interface.
This could bring the G4, at least for a time, up to par with the 970.
Not with the G4 being limited to a 167mhz SDR bus. The G4 is bandwidth starved and upping the clock speed is an exercise in the very definition of "marginal return".
It has everything to do with mosquitos. The disease (which kills something on the order of 5,000 people every single day) is caused by a plasmodium spread by the anopheles mosquito. Endemic malaria is therefore bounded by the home range of the mosquitos that carry it.
It doesn't help that the areas of the world where the disease is endemic tend to be places where there is no history of a robust public health infrastructure (Brazil, West Africa, Southeast Asia). There were widespread spraying programs throughout the '60s that cut the rate of infection dramatically, but there's no political will -- nor, to be fair, sufficient resources -- to keep the program going. Uganda (or possibly Kenya, I forget) has been very successful with the simplest of malaria control measures: insecticide-treated mosquito nets, which cut infection rates dramatically.
Malaria is so brutal because it not only kills, but also debilitates an additional 500,000,000 people a year, who can't do anything -- can't work, can't care for children or sick family, anything. It's not for nothing that malaria is increasingly being viewed as an issue not just of public health, but also of economic development.
A fair argument could even be made that malaria (in addition to schistosomiasis and trypanosomiasis) basically prevented the formation of cities beyond a certain population density. Check out William MacNeill's Plagues And Peoples for an excellent treatment of the impact of disease on human cultural development.
We run a smallish (2.4gb on disk, ~90 tables, a couple of 40,000,000 rows) postgres instance (actually, three, with master/slave replication and select multiplexing), and it works great. We only see about 80 transactions a minute, though, so our needs aren't really that impressive.
Those are good points; however, they're not specific to Sun. Any real corporate support works likes that -- it's just a matter of how much you're willing to spend. As the costs associated with good support radically outweigh those for hardware, I suppose one could argue that it doesn't really matter what the hardware is, as long as you've got this fantastic support organization underneath it.
I have heard -- no, not on slashdot -- that Sun's support organization is slipping even from where it was in the late '90s, when I had a very very bad experience with Gold level support (an incompetent and unprepared tech killed a UE6000, thankfully a hot spare). I know that their quality control on the hardware has slipped dramatically.
The world will become very interesting for Sun once the Opteron and the 970 are widely available.
Slashdot Mirror
Of the things you listed, Oracle 9i is available on OS X; I'll give you Coldfusion; and Veritas isn't a network service, it's a freaking filesystem. I might as well point out that there's not a Linux box out there that can integrate into a netinfo domain -- while technically true, it doesn't matter in even the tiniest bit for network service provision.
'jfb
Linux on the server requires maitnence sure .... but it is also capable of doing/running a crapload more stuff than macOS
... ?
Such as
'jfb
But I think the best quote I ever heard on the subject was: "Macs are only more expensive if your time has no value."
A paraphrase of jwz's "linux is only free if your time has no value." True in both cases.
'jfb
In this case, hashing isn't going to discover anything at all interesting, with or without regular expressions. You could tokenize the source and build a parse tree (big fun with C! *pukes*), or taking a page from the information theory folks, use gzip, and, picking an arbitrary threshold of similarity, visually inspect the suspect chunks.
'jfb
MD5 is goofy stupid for this kind of thing. You'd want to use gzip.
'jfb
January 16, 2003?
It's a cool story, but about six months late, eh?
'jfb
I was, in fact.
'jfb
All three of which are wildly deserving of remakes.
'jfb
Sun's chips are impressive but Intel's economy of scale kills them in price and comes close enough in performance that x86 is the smart decision for 99% of people today.
Good stuff, but one quibble: Sun's chips haven't been impressive, relative to their competition, since 1994. Everybody and their dog walks all over SPARC these days, and Intel is actually the absolute fastest, from P4 for 32bits to Itanic 2 for 64.
The only reason to spend the money on Solaris is if you're stuck with some tedious garbage like SAP. As DEC could tell you, that's not a good way to go about finding new customers. I give Sun two years.
'jfb
Damn tootin'. My girlfriend owns a video store and I don't make enough dough to support us both.
'jfb
You're better off with TERM set to dtterm, if'n your application supports it. Just in case you thought progress in computing had been made in the last 25 years, along comes a happy reminder like terminal handling to show you the error of your ways.
*scowl*
'jfb
Terminal.app also emulates DEC's tres-cool dtterm. Just set your TERM variable to dtterm. It's a superset of xterm-color functionality, as near as I can tell. As it happens, you can make Terminal.app do all kinds of wacky things with escapes. For instance:
alias dock='echo -n "[2t"'
alias lower='echo -n "[6t"'
alias raise 'echo -n "[5t"'
Here're some more.
'jfb
No joke. Dave Pajo might be the most pathetic of the bunch.
'jfb
PowerQuiccIII/MPC8560 supports not only DDR memory, but two Ghz ethernet controllers and a RapidIO controller.
... and no Altivec. It's not a G4 (MPC75xx family), it's a different class of processor entirely. That it implements Book E is of no consequence -- and besides, even with it's (single-channel) DDR, it's still slower and offers still less memory bandwidth than the 970.
'jfb
The interesting thing was that raw number-crunching power wasn't always the most important thing -- many times it is bandwith ...
... which points out in stark and amusing relief the differing worlds of desktop and embedded usage. The reason that Apple is so hot for the 970 is that it offers dual channel DDR memory -- quite a change from the 133mhz SDR that the G4 is limited to.
'jfb
Well, you're half right. The POWER4 is designed for reliability. However, at 1.7ghz, it handily outperforms the projected numbers for a 970, particularly and unsurprisingly in floating point performance.
SPEC2000
POWER4 @ 1.7ghz: 1113/1699 (int/fp)
PPC970 @ 1.8ghz: 937/1085 (int/fp) *projected
Don't get me wrong: as soon as a Mac with this baby in it is available, I'm upgrading, but let's call a spade a spade. The 970 looks to be decently faster than what we currently have in raw processing power, but with a radical, "holy cow where're my pants" faster memory interface.
'jfb
This could bring the G4, at least for a time, up to par with the 970.
Not with the G4 being limited to a 167mhz SDR bus. The G4 is bandwidth starved and upping the clock speed is an exercise in the very definition of "marginal return".
'jfb
If you're hiking at 30,000 ft over sea level, you've got more to worry about than your cell phone reception.
'jfb
Are the Centers For Disease Control And Prevention reliable enough?
'jfb
It has everything to do with mosquitos. The disease (which kills something on the order of 5,000 people every single day) is caused by a plasmodium spread by the anopheles mosquito. Endemic malaria is therefore bounded by the home range of the mosquitos that carry it.
It doesn't help that the areas of the world where the disease is endemic tend to be places where there is no history of a robust public health infrastructure (Brazil, West Africa, Southeast Asia). There were widespread spraying programs throughout the '60s that cut the rate of infection dramatically, but there's no political will -- nor, to be fair, sufficient resources -- to keep the program going. Uganda (or possibly Kenya, I forget) has been very successful with the simplest of malaria control measures: insecticide-treated mosquito nets, which cut infection rates dramatically.
Malaria is so brutal because it not only kills, but also debilitates an additional 500,000,000 people a year, who can't do anything -- can't work, can't care for children or sick family, anything. It's not for nothing that malaria is increasingly being viewed as an issue not just of public health, but also of economic development.
A fair argument could even be made that malaria (in addition to schistosomiasis and trypanosomiasis) basically prevented the formation of cities beyond a certain population density. Check out William MacNeill's Plagues And Peoples for an excellent treatment of the impact of disease on human cultural development.
'jfb
We run a smallish (2.4gb on disk, ~90 tables, a couple of 40,000,000 rows) postgres instance (actually, three, with master/slave replication and select multiplexing), and it works great. We only see about 80 transactions a minute, though, so our needs aren't really that impressive.
'jfb
Those are good points; however, they're not specific to Sun. Any real corporate support works likes that -- it's just a matter of how much you're willing to spend. As the costs associated with good support radically outweigh those for hardware, I suppose one could argue that it doesn't really matter what the hardware is, as long as you've got this fantastic support organization underneath it.
I have heard -- no, not on slashdot -- that Sun's support organization is slipping even from where it was in the late '90s, when I had a very very bad experience with Gold level support (an incompetent and unprepared tech killed a UE6000, thankfully a hot spare). I know that their quality control on the hardware has slipped dramatically.
The world will become very interesting for Sun once the Opteron and the 970 are widely available.
Best,
'jfb
*rimshot*
'jfb
BUT a SUN is a SUN, did you ever wonder why is there a 2K$ gap between the lowest end Sparc and the highest end x86 machine of the same class ?
Because people in purchasing and management are stupid. There hasn't been a Sun worth the premium since the SS20.
'jfb
"For every complex problem, there is an answer that is clear, simple, and wrong." -- HL Mencken
'jfb