Didn't bug me that much, coz I usually wanted to do some useful stuff before the loop, like make persistent database connections, initialize caches, set up logging to syslog, set up other stuff.
The mod_perl stuff tries to do the persistent DB stuff but in a kludgy untidy way that has a lot more gotchas. Same for PHP's mysql_pconnect. Go see people say "turn off persistent DB connections" in one answer and then "turn on persistent DB connections" in another answer;).
I've tried mod_perl, FastCGI is cleaner, less buggy and less bug prone.
On Linux and *BSD you can even have the main fastcgi process fork multiple children and have the children make the persistent connections etc, then all accept on the same fastcgi socket - it still works. Used to be get more shared mem this way than if Apache forked each one separately, but maybe nowadays you don't gain anything from that?
If you don't need total control over how Apache behaves, use mod_fastcgi not mod_perl. For fastcgi typically the webserver forks off the webapp as a separate _persistent_ process and then they talk to each other via a socket (you could actually have the webapp on a separate machine, but I haven't had much need to do that yet). Fastcgi apps can be coded similar to CGI/mod_perl apps there's just a loop around the "Fastcgi accept" call.
With fastcgi you can use perl, python, ruby, C++ or whatever - just like you can use these with CGI.
It may be _slightly_ slower - two processes talking to each other rather than one big process, but it is faster than CGI. And it's more robust and clean than mod_perl.
Also, if apache ever gets too slow, you can switch to Zeus, and you can still use fastcgi there. No change to your code.
Whether you are good or not, is not dependent on whether something deserves your act of goodness or not. Similarly for fairness. It's part of having integrity.
I would think it would be better in the long term to choose to be good, rather than choose to not be evil (ala Google[1]).
[1] Seems Google's policy is to try to "hire above the mean" (which is a good idea for them), but aiming for "don't be evil" is not aiming "above the mean".
You better ensure that average time isn't too highly weighted.
Otherwise the popular solution could be "replace drive with one with a pristine image", "reimage old drive if old drive works".
Alternatively there could be one of those hardware/software solutions that can rollback to a "pristine snapshot". Reboot machine, select "rollback" during the BIOS bootup, and voila, fresh installed system with the usual patches.
Being immortal AND NOT invincible - still vulnerable to pain etc ala Highlander could be a bad thing.
Imagine if you were trapped in a landslide, and you didn't die. You had to spend the next few weeks or even months digging yourself out whilst suffering hunger, pain, thirst etc.
Or worse, some sadistic psycho captures you and finds out that you can't die...
So if you ever had a chance at immortality you better read the fine print and think very carefully.
Alternative to conventional emergency response?
on
Treating the Dead
·
· Score: 1
Actually this may mean you can save a lot more people, not just heart attack and drowning ones. And could change what it means to "stabilize" a person in an emergency.
A single motorcycle can carry a lot of "hypothermia" inducing stuff. You send one or two bikers to the scene instead of one ambulance - bikes can get through gridlock a lot faster[1].
They get there ASAP, put the victim on "pause", and then they have a lot more time to bring them back to the hospital, where the ER team can fix them based on the new knowledge.
You might be able to quickly convert an "ambulance" bike to a tricycle to carry the patient back (while "paused").
[1] Of course one must take measures to significantly reduce the odds of the bikers becoming accident victims themselves;).
Re:This makes sense in a lot of ways.
on
Treating the Dead
·
· Score: 1
AFAIK there are _long_ term negative effects of general anesthetics on brains, at least those of rats[1].
Does the hypothermia method involve the use of GA?
Anyway it would be good to do a comparison (starting with rats I suppose;) ) of GA+ hypothermia method, GA + conventional methods, and just the hypothermia alone without GA.
[1] "Long-term impairment of acquisition of a spatial memory task following isoflurane-nitrous oxide anesthesia in rats."
Anyway, here's a bit more detail on what I suggest:
Ubuntu or someone comes up with a list of top 7-10 sandbox templates that combined would handle 90% of the apps. And users are allowed to make their own custom templates.
** directories There would be a per machine "Programs" directory, and a per user "Programs" directory. Each user will have a separate "program data" directory for each installed program. A guest/transient program will have it in/tmp and it should be cleaned up automatically.
May wish to have a "program settings" directory if it's not combined with "program data".
** example privileges full administrator sound recording sound playback full screen graphics windowed graphics keyboard input pointer input network access (detailed protocol and ports?) storage/filesystem access (none, nonpermanent, app only, app group, user) etc
Note: some sandboxes would still need to provide read/exec access to some libraries/files - this should also be defined in the template (not easy but oh well).
*** Installation and Running should be different things.
**Install/update The privileges you need to run an installer/updater would be different from running the actual installed program. You could limit basic installers to only be able to _add_ directories to the "Programs" directory, "Program Data" directory (which could be per user or per machine- depending on the installer sandbox template and the user's privileges).
There should also be a Documents directory where apps that run in the "desktop/office app sandbox" can read/write to.
The "safe" installer sandbox template might not need working network/sound/full screen access. After all they should just put the darn files there, and stop connecting back home for registering - give a url for registration if you want.
During the install the user should probably already decide what sandbox template to use when _running_ the installed program (if it's not already set by the install sandbox template).
** Run When a user chooses to run an app/executable/script which isn't already associated with a template, the O/S would prompt the user to choose a sandbox template to use, and give a few other options: e.g.
"Fun Screensaver" requests "Screen saver install privileges" Run this with <dropdown>"safe guest applet" privileges</dropdown>, Cancel + OK buttons. Checkbox: always use this template for this application. Manage sandbox templates.
Screen saver install privileges would install the screen saver using a "screen saver install" sandbox template which would also set the resulting installed program to always be run using the "screen saver" sandbox template (no need to prompt the user again).
safe guest applet = windowed graphics, sound output, normal safe input (no strange grabs of the keyboard/mouse/etc when not in focus), no permanent storage, no access to any external directory, no network access).
There will be other sandbox options in the "dropdown".
At this point the system could also do some signed code checks if relevant.
Yep, you got it, that's what I meant - you should only need to do it once if you're installing a program, with that "always grant XYZ privileges" thingy. If you're running some "run once" stuff it's up to you whether you want to keep granting it access.
I guess I kinda slipped up and didn't mention that apps just request and the user still has to approve.
My excuse is nowadays I'm getting too lazy to go into detail - coz seems lots of people still won't understand if I go into detail of how everything would and should work (there are plenty of details to get right so that "Aunt May" can use it reasonably safely). Those that are able to understand it usually get it and the resulting implications almost immediately.
Just look at all that talk about RBAC, SELinux and systrace. Steve Jobs will throw those "solutions" out and with good reason - they are not fit for normal users.
Even IT gurus should have other things to do than detailed RBAC configuration for every single app they run. Most desktop software out there falls into a few categories.
That's why Vista is pretty disappointing - Microsoft's home directory structure actually would be fine - there's already \Application Data \Local Settings \My Documents and so on.
Worse: there are already plenty of 3rd party windows personal firewalls/security programs that prevent network access in detail on a per app/executable basis etc, that means the hooks are all there.
So all Microsoft needs to do is the template thingy - they have a list of thousands of apps, they should be able to come up with "top 10" templates that would work with most of the popular apps, and then allow users to create custom templates.
The current UAC means you have to keep going through the same decisions over and over again for the security level you want or turn it off for low security.
The template thing I propose means you don't have to keep repeating yourself, and you get to keep the security level you chose (barring any O/S bugs;) ).
A modern OS should be having something that's much better than sudo.
Modern desktop class OSes should have sandbox _templates_, with apps being allowed to "suggest" a template.
Then if an app claims to be a "plain old screen saver", it only gets "plain old screen saver" rights - which means no network access, no access to the user's files etc.
If it claimed to be a "standard network game" then it gets different sort of access - file system access to its own "app specific data folder" in the user's home directory, access to full-screen graphics, sound _playback_ (not recording[1]), limited network access (as per requested).
If some flash applet "game" somehow requires "full administrative system privileges", go figure...
[1] Only a few apps should be allowed to record sound - stuff like skype, voice chat app for games. Your word processor should not be recording sound. The O/S should handle the voice control stuff if you like that sort of crap. And by default you may not wish to allow an app to record sound while backgrounded or just sitting in the "systray".
Seriously though, what's the big deal? Even when people tried to emulate Blizzard's stuff legitimately they still got shut down.
It's an MMOG, the number of useful "gems" in the code are probably really low. You'd probably be able to come up with those "gems" independently anyway.
Also you could just get them to boast about their great new features to the media way before their release and a smart person in your company could figure out how to do it in 5 minutes. You don't have such smart people in your company? Too bad then.
If you left your company you'd probably want to be be writing the _interesting_ nonboilerplate, nonlib[1] stuff from scratch. After all you should be doing things much better plus take advantage of the latest advances in hardware and software the next time round right? Bonus if you don't have to be drop-in backward compatible with the previous crap you wrote:).
Just look at John Carmack. He gives everyone access to his old game engines. Even if someone copied his _latest_ game engine right off his PC, so what? They don't have him.
If the Lineage III source code really is worth 1 billion, then if you can "buy" John Carmack's coding+designing services for 5 years for <= 500 million you've got yourself a great deal eh? No? Why not?
Code doesn't make money just sitting there. There's a LOT of other equally important stuff required. You could have two different MMOGs with the same engines and they can be VERY VERY different with different resulting net profit. Different artwork, sound, music, story, game play, game balance, community management, availability+reliability, etc.
Anyone think millions of people play WoW because the game engine is so good? Doh.
Imagine someone stealing the windows source AND trying to use it to make money without using the windows noncode stuff. The only viable way would probably be for finding security bugs faster.
[1] While the company can have sole rights to what you make for them, I don't recommend you let companies own sole rights to your libraries - that's like a carpenter allowing a company to own his tools just because he tweaked them while making stuff for the company. They can _copy_ my libs if they want, but it is wrong for them to say it's now theirs alone to use just because I changed or even added stuff while I happened to be working for them. If they do that then that really is _theft_.
Calling something a voting machine AND selling it, when it isn't a voting machine fit for that purpose at all.
It's as much a voting machine as my bedsheet is a certified parachute fit for skydivers.
They're lucky they're in the USA. In other countries they might actually get lynched by angry voters or executed (for treason?) if they escape the mob.
But it'll be hard to convince the rest of the world that the US is interested in democracy and fair elections in Iraq if this sort of thing keeps happening in the USA.
Maybe the USA should just modify and use the American Idol voting system. It can't be much worse than using Microsoft Jet and all the other crap Diebold is making.
While the first test just showed it's not good for postgresql, the second test I linked to showed it's not worth it for Apache+PHP+MySQL or java+Sybase.
Sure it's got to be good for some stuff - maybe pure java threaded apps- no db, no apache, no floating point.
BUT it seems too specialized with plenty of performance gotchas so why bother with T1? It's not like it's really cheaper, or has better performance/watt (go look at the results of the 2nd test).
So, please do show me benchmarks where the T1 stomps the opterons and the woodcrests.
Anyone actually think the T2 will be competitive by the time it gets released? If I were still on SPARC I'd be moving off ASAP. It's not like Suns SPARCs are significantly more reliable/available - they're just conventional Unix systems. Not like Tandem or VMS. Even IBM is better for HA unix systems.
Oops. That was the T1, not T2 that was crushed. I don't think the T2 is out yet.
See: http://tweakers.net/reviews/649/10 (2006 October, T1 vs Opteron) Quote: "Unfortunately, we have little choice but to be disappointed in the UltraSparc T1's performance: even the perfectly scaling PostgreSQL allows the machine to be very convincingly overtaken by the 'average' Opteron server, costing just below half its price."
http://www.anandtech.com/printarticle.aspx?i=2772 (2006 June, Intel 5160 vs Opteron vs T1) Quote: Intel Xeon 5160 (Woodcrest) Advantages:
* Best server performance across all applications
* Best Performance/Watt in the high end
* Absolutely stunning web server performance
* FB-DIMM enables high RAM capacity and bandwidth (quad channel)
Sun's SPARC CPUs are like little bugs squished in the Intel vs AMD CPU battles.
So who's gonna bet that the T2 would actually be competitive when it gets released?
If you have all your innodb tables in one big multiGB innodb file (as required by old MySQL4 versions) then that dumping and reloading takes quite a long time.
Slashdot Mirror
It's not even a new problem to solve.
There were plenty of ancient _empires_ that got big by bullying (and worse), and guess how most of those leaders behaved.
Do we really want more of this? They only get more powerful if they can bully others to help them.
You can log off out of real life. The trouble is logging back on again.
;).
Some people say you have to start a totally new character to log on again.
And others say you usually can't log on back again, you have to logon to somewhere else instead.
Don't like it? Go look for the SLAs
s/vagina/anus/ then. Most people have anuses.
;).
Not that I'm actually trying to promote rape or anal sex. But hey this is a "virtual" discussion right?
Didn't bug me that much, coz I usually wanted to do some useful stuff before the loop, like make persistent database connections, initialize caches, set up logging to syslog, set up other stuff.
;).
The mod_perl stuff tries to do the persistent DB stuff but in a kludgy untidy way that has a lot more gotchas. Same for PHP's mysql_pconnect. Go see people say "turn off persistent DB connections" in one answer and then "turn on persistent DB connections" in another answer
I've tried mod_perl, FastCGI is cleaner, less buggy and less bug prone.
On Linux and *BSD you can even have the main fastcgi process fork multiple children and have the children make the persistent connections etc, then all accept on the same fastcgi socket - it still works. Used to be get more shared mem this way than if Apache forked each one separately, but maybe nowadays you don't gain anything from that?
Hey, what if you have some pervert whose fantasy is wanting to have consensual sex with FBI agents who pretend to be teenage girls?
:).
But then what happens if it's a teenage girl who is pretending to be an FBI agent pretending to be a teenage girl?
Maybe it'll be a bit simpler if you have to marry a person before you have sex with that person. Lots of people don't like that approach though
Whoopee, if a 3rd party can forcibly script an avatar to do something, then it's viruses and worms time!
If you don't need total control over how Apache behaves, use mod_fastcgi not mod_perl. For fastcgi typically the webserver forks off the webapp as a separate _persistent_ process and then they talk to each other via a socket (you could actually have the webapp on a separate machine, but I haven't had much need to do that yet). Fastcgi apps can be coded similar to CGI/mod_perl apps there's just a loop around the "Fastcgi accept" call.
With fastcgi you can use perl, python, ruby, C++ or whatever - just like you can use these with CGI.
It may be _slightly_ slower - two processes talking to each other rather than one big process, but it is faster than CGI. And it's more robust and clean than mod_perl.
Also, if apache ever gets too slow, you can switch to Zeus, and you can still use fastcgi there. No change to your code.
Whether you are good or not, is not dependent on whether something deserves your act of goodness or not. Similarly for fairness. It's part of having integrity.
I would think it would be better in the long term to choose to be good, rather than choose to not be evil (ala Google[1]).
[1] Seems Google's policy is to try to "hire above the mean" (which is a good idea for them), but aiming for "don't be evil" is not aiming "above the mean".
You better ensure that average time isn't too highly weighted.
Otherwise the popular solution could be "replace drive with one with a pristine image", "reimage old drive if old drive works".
Alternatively there could be one of those hardware/software solutions that can rollback to a "pristine snapshot". Reboot machine, select "rollback" during the BIOS bootup, and voila, fresh installed system with the usual patches.
Sure the symptoms get fixed quickly, but...
Being immortal AND NOT invincible - still vulnerable to pain etc ala Highlander could be a bad thing.
Imagine if you were trapped in a landslide, and you didn't die. You had to spend the next few weeks or even months digging yourself out whilst suffering hunger, pain, thirst etc.
Or worse, some sadistic psycho captures you and finds out that you can't die...
So if you ever had a chance at immortality you better read the fine print and think very carefully.
Actually this may mean you can save a lot more people, not just heart attack and drowning ones. And could change what it means to "stabilize" a person in an emergency.
;).
A single motorcycle can carry a lot of "hypothermia" inducing stuff. You send one or two bikers to the scene instead of one ambulance - bikes can get through gridlock a lot faster[1].
They get there ASAP, put the victim on "pause", and then they have a lot more time to bring them back to the hospital, where the ER team can fix them based on the new knowledge.
You might be able to quickly convert an "ambulance" bike to a tricycle to carry the patient back (while "paused").
[1] Of course one must take measures to significantly reduce the odds of the bikers becoming accident victims themselves
AFAIK there are _long_ term negative effects of general anesthetics on brains, at least those of rats[1].
;) ) of GA+ hypothermia method, GA + conventional methods, and just the hypothermia alone without GA.
= Retrieve&db=PubMed&list_uids=14739805&dopt=Abstrac t
Does the hypothermia method involve the use of GA?
Anyway it would be good to do a comparison (starting with rats I suppose
[1] "Long-term impairment of acquisition of a spatial memory task following isoflurane-nitrous oxide anesthesia in rats."
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd
Too lazy to sign up to post :p.
/tmp and it should be cleaned up automatically.
Anyway, here's a bit more detail on what I suggest:
Ubuntu or someone comes up with a list of top 7-10 sandbox templates that combined would handle 90% of the apps. And users are allowed to make their own custom templates.
** directories
There would be a per machine "Programs" directory, and a per user "Programs" directory.
Each user will have a separate "program data" directory for each installed program. A guest/transient program will have it in
May wish to have a "program settings" directory if it's not combined with "program data".
** example privileges
full administrator
sound recording
sound playback
full screen graphics
windowed graphics
keyboard input
pointer input
network access (detailed protocol and ports?)
storage/filesystem access (none, nonpermanent, app only, app group, user)
etc
Note: some sandboxes would still need to provide read/exec access to some libraries/files - this should also be defined in the template (not easy but oh well).
*** Installation and Running should be different things.
**Install/update
The privileges you need to run an installer/updater would be different from running the actual installed program. You could limit basic installers to only be able to _add_ directories to the "Programs" directory, "Program Data" directory (which could be per user or per machine- depending on the installer sandbox template and the user's privileges).
There should also be a Documents directory where apps that run in the "desktop/office app sandbox" can read/write to.
The "safe" installer sandbox template might not need working network/sound/full screen access. After all they should just put the darn files there, and stop connecting back home for registering - give a url for registration if you want.
During the install the user should probably already decide what sandbox template to use when _running_ the installed program (if it's not already set by the install sandbox template).
** Run
When a user chooses to run an app/executable/script which isn't already associated with a template, the O/S would prompt the user to choose a sandbox template to use, and give a few other options:
e.g.
"Fun Screensaver" requests "Screen saver install privileges"
Run this with <dropdown>"safe guest applet" privileges</dropdown>, Cancel + OK buttons.
Checkbox: always use this template for this application.
Manage sandbox templates.
Screen saver install privileges would install the screen saver using a "screen saver install" sandbox template which would also set the resulting installed program to always be run using the "screen saver" sandbox template (no need to prompt the user again).
safe guest applet = windowed graphics, sound output, normal safe input (no strange grabs of the keyboard/mouse/etc when not in focus), no permanent storage, no access to any external directory, no network access).
There will be other sandbox options in the "dropdown".
At this point the system could also do some signed code checks if relevant.
Yep, you got it, that's what I meant - you should only need to do it once if you're installing a program, with that "always grant XYZ privileges" thingy. If you're running some "run once" stuff it's up to you whether you want to keep granting it access.
;) ).
I guess I kinda slipped up and didn't mention that apps just request and the user still has to approve.
My excuse is nowadays I'm getting too lazy to go into detail - coz seems lots of people still won't understand if I go into detail of how everything would and should work (there are plenty of details to get right so that "Aunt May" can use it reasonably safely). Those that are able to understand it usually get it and the resulting implications almost immediately.
Just look at all that talk about RBAC, SELinux and systrace. Steve Jobs will throw those "solutions" out and with good reason - they are not fit for normal users.
Even IT gurus should have other things to do than detailed RBAC configuration for every single app they run. Most desktop software out there falls into a few categories.
That's why Vista is pretty disappointing - Microsoft's home directory structure actually would be fine - there's already \Application Data \Local Settings \My Documents and so on.
Worse: there are already plenty of 3rd party windows personal firewalls/security programs that prevent network access in detail on a per app/executable basis etc, that means the hooks are all there.
So all Microsoft needs to do is the template thingy - they have a list of thousands of apps, they should be able to come up with "top 10" templates that would work with most of the popular apps, and then allow users to create custom templates.
The current UAC means you have to keep going through the same decisions over and over again for the security level you want or turn it off for low security.
The template thing I propose means you don't have to keep repeating yourself, and you get to keep the security level you chose (barring any O/S bugs
A modern OS should be having something that's much better than sudo.
Modern desktop class OSes should have sandbox _templates_, with apps being allowed to "suggest" a template.
Then if an app claims to be a "plain old screen saver", it only gets "plain old screen saver" rights - which means no network access, no access to the user's files etc.
If it claimed to be a "standard network game" then it gets different sort of access - file system access to its own "app specific data folder" in the user's home directory, access to full-screen graphics, sound _playback_ (not recording[1]), limited network access (as per requested).
If some flash applet "game" somehow requires "full administrative system privileges", go figure...
[1] Only a few apps should be allowed to record sound - stuff like skype, voice chat app for games. Your word processor should not be recording sound. The O/S should handle the voice control stuff if you like that sort of crap. And by default you may not wish to allow an app to record sound while backgrounded or just sitting in the "systray".
You can sit in country A while the software you wrote is compiled and installed in country B where the patent laws don't apply.
They should have written it in perl. ;).
:).
Seriously though, what's the big deal? Even when people tried to emulate Blizzard's stuff legitimately they still got shut down.
It's an MMOG, the number of useful "gems" in the code are probably really low. You'd probably be able to come up with those "gems" independently anyway.
Also you could just get them to boast about their great new features to the media way before their release and a smart person in your company could figure out how to do it in 5 minutes. You don't have such smart people in your company? Too bad then.
If you left your company you'd probably want to be be writing the _interesting_ nonboilerplate, nonlib[1] stuff from scratch. After all you should be doing things much better plus take advantage of the latest advances in hardware and software the next time round right? Bonus if you don't have to be drop-in backward compatible with the previous crap you wrote
Just look at John Carmack. He gives everyone access to his old game engines. Even if someone copied his _latest_ game engine right off his PC, so what? They don't have him.
If the Lineage III source code really is worth 1 billion, then if you can "buy" John Carmack's coding+designing services for 5 years for <= 500 million you've got yourself a great deal eh? No? Why not?
Code doesn't make money just sitting there. There's a LOT of other equally important stuff required. You could have two different MMOGs with the same engines and they can be VERY VERY different with different resulting net profit. Different artwork, sound, music, story, game play, game balance, community management, availability+reliability, etc.
Anyone think millions of people play WoW because the game engine is so good? Doh.
Imagine someone stealing the windows source AND trying to use it to make money without using the windows noncode stuff. The only viable way would probably be for finding security bugs faster.
[1] While the company can have sole rights to what you make for them, I don't recommend you let companies own sole rights to your libraries - that's like a carpenter allowing a company to own his tools just because he tweaked them while making stuff for the company. They can _copy_ my libs if they want, but it is wrong for them to say it's now theirs alone to use just because I changed or even added stuff while I happened to be working for them. If they do that then that really is _theft_.
Well maybe that's how the 1 minute per person doctors have more time to spend on golf, writing webcam drivers, kernel hacking etc ;).
Isn't this fraud in so many ways?
Calling something a voting machine AND selling it, when it isn't a voting machine fit for that purpose at all.
It's as much a voting machine as my bedsheet is a certified parachute fit for skydivers.
They're lucky they're in the USA. In other countries they might actually get lynched by angry voters or executed (for treason?) if they escape the mob.
But it'll be hard to convince the rest of the world that the US is interested in democracy and fair elections in Iraq if this sort of thing keeps happening in the USA.
Maybe the USA should just modify and use the American Idol voting system. It can't be much worse than using Microsoft Jet and all the other crap Diebold is making.
While the first test just showed it's not good for postgresql, the second test I linked to showed it's not worth it for Apache+PHP+MySQL or java+Sybase.
Sure it's got to be good for some stuff - maybe pure java threaded apps- no db, no apache, no floating point.
BUT it seems too specialized with plenty of performance gotchas so why bother with T1? It's not like it's really cheaper, or has better performance/watt (go look at the results of the 2nd test).
So, please do show me benchmarks where the T1 stomps the opterons and the woodcrests.
Anyone actually think the T2 will be competitive by the time it gets released? If I were still on SPARC I'd be moving off ASAP. It's not like Suns SPARCs are significantly more reliable/available - they're just conventional Unix systems. Not like Tandem or VMS. Even IBM is better for HA unix systems.
ORLY? Junior staff get that much authority? Such empowerment.
;).
Maybe it's an ex-senior staffer who's now a junior staffer or is that even more unbelievable nowadays
Oops. That was the T1, not T2 that was crushed. I don't think the T2 is out yet.
See:
http://tweakers.net/reviews/649/10
(2006 October, T1 vs Opteron)
Quote: "Unfortunately, we have little choice but to be disappointed in the UltraSparc T1's performance: even the perfectly scaling PostgreSQL allows the machine to be very convincingly overtaken by the 'average' Opteron server, costing just below half its price."
http://www.anandtech.com/printarticle.aspx?i=2772
(2006 June, Intel 5160 vs Opteron vs T1)
Quote:
Intel Xeon 5160 (Woodcrest)
Advantages:
* Best server performance across all applications
* Best Performance/Watt in the high end
* Absolutely stunning web server performance
* FB-DIMM enables high RAM capacity and bandwidth (quad channel)
Sun's SPARC CPUs are like little bugs squished in the Intel vs AMD CPU battles.
So who's gonna bet that the T2 would actually be competitive when it gets released?
If you have all your innodb tables in one big multiGB innodb file (as required by old MySQL4 versions) then that dumping and reloading takes quite a long time.
Too long.
I think it usually helps :).
You mispelled weird. ;)