* broken drivers, signed or unsigned, will always cause hardware issues. But then, I don't really have a fundamental problem with drivers signed by Microsoft just as much as I like 'certified hardware' to run Suse or Redhat on.
* the wintel CPU has 4 protection rings - but as far as I can tell "modern OSes" only use two of them. If you're after mainframe style security is VMS good enough?
There's a big difference here. Microsoft are deciding, on your behalf, what you should be running.
The thing that really disgusts me, besides clicking the X and shutting down the Safari tab whilst trying to do some research, is that its bypassing a fundamental problem with software and 'desktop operating systems' - there really is too much trust.
Microsofts proposal doesn't fix the software to be more paranoid - its simply preventing the software from running in the first place.
It really really digusts me that something embedded in email can send out email. Very much.
Now, I have a vague recollection of something which I'd like an intrepid slashdot reader to find. It involved granting _libraries_ permissions to certain resources.
Imagine, for a moment, that you granted the libicmp library the only permission to the ICMP socket. When you linked to the libicmp library you'd be given specific entry points. Its now up to the library to verify that you're able to grab the ICMP socket - and it can also limit how often, how large, etc.
(If you still don't get it - think SUID-style libraries.)
Does anyone recall where this came from? I could have sworn I saw it in a Multics-related post.
.. you sound over 18. Or, at least, mature enough to understand the difference between make believe and reality. To see _past_ the simple parallels you could draw between real life situations and depicted violence in the game.
Wrong. They are in real mode but its not in 4.77mhz mode. Besides, it'd be hard to emulate the 8-bit external datapath..:)
In other news - underclocking a CPU isn't always as easy as one would think. The internal timings of the CPU are quite tightly meshed with, say, the decay rate on various capacitor-based things (perhaps like cache, registers, etc?).
Otherwise I'd be buying an AMD-1800xp, clocking it at 600mhz or something and having it real cool. AMD - if you can, at all, make a home CPU which you can underclock to the point where a fan isn't required you might find some marketshare in smaller/server devices.
Case in point. The University of Western Australia, Computer Science degree _does not require_ Math. At all. I'm serious. I _chose_ to do first year math.
Admittedly I dropped out after first semester, again, out of sheer boredom.
First semester: Java, Haskell, IT Foundations (word, excel, javascript).
Second Semester: Matlab (image processing stuff), some UML IIRC.
At least this years' IT Foundations course is based around Linux rather than Windows. Go mr Glance!
This illustrates a point - CS courses these days, at least from what I've seen here in Australia, are very job-oriented. You can learn actual CS theory stuff if you choose to but the CS courses that I've seen lead you into multimedia, IT, Application design..
(Personal note: the univeristy of sydney, at least, has an advanced course set for clueful students but I'm on the wrong side of the country for it. After I finish second semester _this_ year I'll take my completed first year and move into something completely different as CS doesn't really kick it for me..)
You've obviously never written a system that had to be audited.
A well thought out database with RI/foreign keys, triggers and stored procedures can be a _joy_ to work with.
Thats the whole idea of triggers and stored procedures - if they're annoying you then you're either doing something wrong or the _integrity_ of the database hasn't been well thought out.
Push some of the application verification into the SQL server for integrity and leave the perl code for UI.
My main beef here is that the majority of web developers see the database as just a storage space and their front end application code (in perl, php, C++, C, ejb, whatever) as the interface. What do I do if I wish to extend the application by adding bits of my own code whilst not breaking? Speak XML to perl code which arbitrates all of my SQL calls? Or speak stored procedures? Hmm.
Case in Point: an ISP I work for uses a system which implements the business logic (stuff like "add x cents to this account based on how much traffic at what rate) into a big precompiled Java library. Cool. Now I want to extend things a little by importing netflow data. I don't have much of a choice how I do it - I _have_ to use their libraries. Which are slow, cause the database to lock up and require me to start up a 200meg java vm just to import a _single_ accounting record. If it were implemented, say, in postgres w/ stored procedures the API could be "just call USER_ADD_TRAFFIC_INFO(userid, timestamp, len, trafin, trafout)".
And for those of you who will say "but what if I want to have an external event! like sending an email!" - use a queue. Have an event queue which is polled by your application. Heck, Postgres _can_ send email out from a stored procedure (there's C and Perl bindings IIRC but I only use plpgsql..)
dot-array LED signs are generally built using some form of scanning setup - you enable a row, then shift-on the bits. Then, next row, shift on the bits, etc. Not all the LEDs are on at any given time.
So you can get away with pulsing them with higher current than they're rated at. Which is a bit of a bugger - if you hit the 'stop' button or the sign crashed, any on LEDs would burn out. I used to work making LED signs for a little while - we had a set of damaged signs to test code on.
(Which got real expensive when Blue LEDs came out so I _think_ the guys implemented some 'no clock? no driving power!' circuitry in case the testing code crashed..:)
(A cute tidbit: our signs had 4-LED RGB elements - one blue, one green, two red. the red leds weren't as bright to the eye, so we needed two of em..)
Oh, I agree. But its not 'rich' as much as 'can absorb cost of purchasing kit over a few years'.
Cisco are also well known for offering quite substantial discounts - but their top end kit is still very expensive.
The cute thing to note is that Cisco, although expensive, do quite a lot of R&D (they'd have to in order to stay above the curve.)
The other cute thing to note is that Nortel, a very well known name in the Telephony equipment (think voice switches) tried the same price tactics with their IP equipment only to find noone kept using them. Having to buy a $20,000 minimum yearly support contract before they'll even _help_ you with a bug in their software is just outrageous.
A nice point for cisco, whilst I'm ranty - they have _excellent_ online documentation. It took me three months to wade through the 20,000+ odd pages of Passport 74xx switch documentation in order to figure out how to configure stuff up. Cisco OTOH have an abundance of example configurations on their website. Its very useful.
The Z80 did have an abundance of registers and the block moves were quite useful and fast.
And the hidden instructions! when you found you could treat IX and IY as 8-bit registers (XH, XL, YH, YL) made for some fast sprite routines.
Oh, and the alternate register set was very useful when writing little interrupt handlers - you could just swap register sets to save push/pop'ing everything..
Yes, but the 6510 had this cute thing called "zero-page" which meant you had quite a bit of scratch space if you knew what you were doing. It was faster than doing full 16-bit address operations.
The nicest thing that I find about working on/with BSD software is that you _can_ keep it proprietary - or could come to the realisation that if you release bits and pieces of it to the public you'll end up having them improved.
If everyone kept their BSD improvements to themselves then we wouldn't be advancing anywhere nearly as fast.
Case in point - there's a netgraph l2tp module. You can combine them with other netgraph modules to have a complete in-kernel l2tp->ppp->ip encaps and decaps path. Very fast. However, there's no userland implementation that exists. This leads me to think that someone _wrote_ an l2tpd, released the ng_l2tp module into the kernel and hoped that people would improve on it. The community benefits from having an l2tp module written, the company benefits from any improvements to it.
Australia also has a plentiful supply of Natural Gas which petrol stations are (mostly, if not all) fitted to supply.
Unfortunately it seems that LPG (Liquefied Petroleum Gas) prices fluctuate just as much as petrol prices, negating any real benefit from moving over to LPG.
Rumours from my car-smart friends outline that an engine built to _just_ run on LPG would be much, much more efficient than an engine converted to run on both unleaded and LPG - does anyone know of any studies/prototypes outlining engine power/efficiency from LPG and how suitabled it'd be for a general-purpose car?
Changing the page granularity just means you need less page table entries in your OS.
Eg, if you have 4gb of ram in your machine and you're using it all, calculate how many PTEs there are.
Now, say you're running Oracle which, IIRC (and I haven't ever run Oracle, but I've done similar tricks with other bits of software) mmap()s a big ass buffer between multiple processes, if you map 1gb of RAM thats 262,144 PTEs at 4k pages. Over, say, 30 processes thats.. lots of PTEs.
If you flip the size bit and get the larger pages you just reduce the PTE size. The main factor here is that the external address stuff is _still_ 32 bit.
Now, PAE lets you do some bank swapping magic _WHICH_ can be useful for things like disk buffers. But for running code and mmap()ed buffers things look tricky.
You have a file (/etc/rc.conf) which specifies configuration parameters for the various _system_ programs.
Under FreeBSD, ports are started out of/usr/local/etc/rc.d/ in a vaguely sysv-init-script fashion (start/stop).
Personally I find the IRIX chkconfig system quite nice. The problem with BSD is that it doesn't extend to any addon software. The problem with normal SYSV is that you rename/modify the scripts - which are then un-renamed or replaced when it comes time to upgrade. I've suffered many hours of frustration at/etc/init.d/sendmail and its vairous links under Solaris after installing a patchset..
(For those who don't know about chkconfig - it maintains a configuration fileset in/var/config/ (symlink to/etc/config? My IRIX box isn't reachable from here!). So/var/config/${programname} can be accessed from chkconfig ${programname}. So, you just write all your/etc/init.d/ scripts to run chkconfig to return whether a program is meant to run or not. You don't ever have to fiddle with the symlinks in/etc/rc?.d/ to change whether stuff starts or not.
Do you, shock, grow your own food? Or buy it from the supermarket?
There's much more you can do for the environment and general well-being of the planet than worrying about the cruelty and inefficincy of animal products.
Search the RFC/IETF sites for URs or Universal Resource Names.
Squid implemented some basic URN stuff a while back - It'd run a cgi script at a URL which returned a list of sites in a specific format. Squid would then ping the sites and return them to you w/ the RTTs. You could then choose the 'best one'.
Not very effective - and I don't think it (the squid side) ever progressed past the initial stages - but maybe its the beginning for someone to tackle this issue again.
(IMHO it could be made more.. transparent. But then this requires caching servers everywhere and thats a thread I don't really want to get into right now.)
Note - there's RFCs covering URNs for various things (IETF documents, ISBNs, etc)..
No, DNS isn't really designed for this. Yes, DNS has been perverted enough, so I guess this'd just be another extension. Write an RFC, get it published..;-)
Slashdot Mirror
Its the speed of light in a vacuum which counts.
But, from the company shareholders (public or private), did he do the right thing?
The right thing is "make money for shareholders". Did he manage to do this?
Sure, it might have been short-term gain rather than long-term prosperity but not all companies are up for the risk that long-term gambles present.
Besides, with the way you've presented the way events occured public companies might love the guy.
* broken drivers, signed or unsigned, will always cause hardware issues. But then, I don't really have a fundamental problem with drivers signed by Microsoft just as much as I like 'certified hardware' to run Suse or Redhat on.
* the wintel CPU has 4 protection rings - but as far as I can tell "modern OSes" only use two of them. If you're after mainframe style security is VMS good enough?
There's a big difference here. Microsoft are deciding, on your behalf, what you should be running.
The thing that really disgusts me, besides clicking the X and shutting down the Safari tab whilst trying to do some research, is that its bypassing a fundamental problem with software and 'desktop operating systems' - there really is too much trust.
Microsofts proposal doesn't fix the software to be more paranoid - its simply preventing the software from running in the first place.
It really really digusts me that something embedded in email can send out email. Very much.
Now, I have a vague recollection of something which I'd like an intrepid slashdot reader to find. It involved granting _libraries_ permissions to certain resources.
Imagine, for a moment, that you granted the libicmp library the only permission to the ICMP socket. When you linked to the libicmp library you'd be given specific entry points. Its now up to the library to verify that you're able to grab the ICMP socket - and it can also limit how often, how large, etc.
(If you still don't get it - think SUID-style libraries.)
Does anyone recall where this came from? I could have sworn I saw it in a Multics-related post.
.. you sound over 18. Or, at least, mature enough to understand the difference between make believe and reality. To see _past_ the simple parallels you could draw between real life situations and depicted violence in the game.
Wrong. They are in real mode but its not in 4.77mhz mode. :)
Besides, it'd be hard to emulate the 8-bit external datapath..
In other news - underclocking a CPU isn't always as easy as one would think. The internal timings of the CPU are quite tightly meshed with, say, the decay rate on various capacitor-based things (perhaps like cache, registers, etc?).
Otherwise I'd be buying an AMD-1800xp, clocking it at 600mhz or something and having it real cool. AMD - if you can, at all, make a home CPU which you can underclock to the point where a fan isn't required you might find some marketshare in smaller/server devices.
Yup and that badly-written .net or java application which you converted to with your thin-client setup will need most of those pennies.
.. Do you remember what this membrane is? :)
Some further information would be lovely.
Perhaps - or perhaps Y! host some servers in Canada and they haven't caught up with the news being originated most probably from the US.
"one website" != "one server" or "servers at one location" these days.
Case in point. The University of Western Australia, Computer Science degree _does not require_ Math. At all. I'm serious. I _chose_ to do first year math.
Admittedly I dropped out after first semester, again, out of sheer boredom.
First semester: Java, Haskell, IT Foundations (word, excel, javascript).
Second Semester: Matlab (image processing stuff), some UML IIRC.
At least this years' IT Foundations course is based around Linux rather than Windows. Go mr Glance!
This illustrates a point - CS courses these days, at least from what I've seen here in Australia, are very job-oriented. You can learn actual CS theory stuff if you choose to but the CS courses that I've seen lead you into multimedia, IT, Application design..
(Personal note: the univeristy of sydney, at least, has an advanced course set for clueful students but I'm on the wrong side of the country for it. After I finish second semester _this_ year I'll take my completed first year and move into something completely different as CS doesn't really kick it for me..)
You've obviously never written a system that had to be audited.
A well thought out database with RI/foreign keys, triggers and stored procedures can be a _joy_ to work with.
Thats the whole idea of triggers and stored procedures - if they're annoying you then you're either doing something wrong or the _integrity_ of the database hasn't been well thought out.
Push some of the application verification into the SQL server for integrity and leave the perl code for UI.
My main beef here is that the majority of web developers see the database as just a storage space and their front end application code (in perl, php, C++, C, ejb, whatever) as the interface. What do I do if I wish to extend the application by adding bits of my own code whilst not breaking? Speak XML to perl code which arbitrates all of my SQL calls? Or speak stored procedures? Hmm.
Case in Point: an ISP I work for uses a system which implements the business logic (stuff like "add x cents to this account based on how much traffic at what rate) into a big precompiled Java library. Cool. Now I want to extend things a little by importing netflow data. I don't have much of a choice how I do it - I _have_ to use their libraries. Which are slow, cause the database to lock up and require me to start up a 200meg java vm just to import a _single_ accounting record. If it were implemented, say, in postgres w/ stored procedures the API could be "just call USER_ADD_TRAFFIC_INFO(userid, timestamp, len, trafin, trafout)".
And for those of you who will say "but what if I want to have an external event! like sending an email!" - use a queue. Have an event queue which is polled by your application. Heck, Postgres _can_ send email out from a stored procedure (there's C and Perl bindings IIRC but I only use plpgsql..)
how high?
:)
dot-array LED signs are generally built using some form of scanning setup - you enable a row, then shift-on the bits. Then, next row, shift on the bits, etc. Not all the LEDs are on at any given time.
So you can get away with pulsing them with higher current than they're rated at. Which is a bit of a bugger - if you hit the 'stop' button or the sign crashed, any on LEDs would burn out. I used to work making LED signs for a little while - we had a set of damaged signs to test code on.
(Which got real expensive when Blue LEDs came out so I _think_ the guys implemented some 'no clock? no driving power!' circuitry in case the testing code crashed..
(A cute tidbit: our signs had 4-LED RGB elements - one blue, one green, two red. the red leds weren't as bright to the eye, so we needed two of em..)
Oh, I agree. But its not 'rich' as much as 'can absorb cost of purchasing kit over a few years'.
Cisco are also well known for offering quite substantial discounts - but their top end kit is still very expensive.
The cute thing to note is that Cisco, although expensive, do quite a lot of R&D (they'd have to in order to stay above the curve.)
The other cute thing to note is that Nortel, a very well known name in the Telephony equipment (think voice switches) tried the same price tactics with their IP equipment only to find noone kept using them. Having to buy a $20,000 minimum yearly support contract before they'll even _help_ you with a bug in their software is just outrageous.
A nice point for cisco, whilst I'm ranty - they have _excellent_ online documentation. It took me three months to wade through the 20,000+ odd pages of Passport 74xx switch documentation in order to figure out how to configure stuff up. Cisco OTOH have an abundance of example configurations on their website. Its very useful.
Or the cost of the equipment? has anyone here seen, even after heavy discounts, the price of equipment to route >OC3 rates of traffic?
The Z80 did have an abundance of registers and the block moves were quite useful and fast.
And the hidden instructions! when you found you could treat IX and IY as 8-bit registers (XH, XL, YH, YL) made for some fast sprite routines.
Oh, and the alternate register set was very useful when writing little interrupt handlers - you could just swap register sets to save push/pop'ing everything..
Yes, but the 6510 had this cute thing called "zero-page" which meant you had quite a bit of scratch space if you knew what you were doing. It was faster than doing full 16-bit address operations.
3 registers isn't nearly as bad as you think.
Right. Tell that to all of the BSDs.
The nicest thing that I find about working on/with BSD software is that you _can_ keep it proprietary - or could come to the realisation that if you release bits and pieces of it to the public you'll end up having them improved.
If everyone kept their BSD improvements to themselves then we wouldn't be advancing anywhere nearly as fast.
Case in point - there's a netgraph l2tp module. You can combine them with other netgraph modules to have a complete in-kernel l2tp->ppp->ip encaps and decaps path. Very fast. However, there's no userland implementation that exists. This leads me to think that someone _wrote_ an l2tpd, released the ng_l2tp module into the kernel and hoped that people would improve on it. The community benefits from having an l2tp module written, the company benefits from any improvements to it.
Its called "mutual benefit". I love it.
Australia also has a plentiful supply of Natural Gas which petrol stations are (mostly, if not all) fitted to supply.
Unfortunately it seems that LPG (Liquefied Petroleum Gas) prices fluctuate just as much as petrol prices, negating any real benefit from moving over to LPG.
Rumours from my car-smart friends outline that an engine built to _just_ run on LPG would be much, much more efficient than an engine converted to run on both unleaded and LPG - does anyone know of any studies/prototypes outlining engine power/efficiency from LPG and how suitabled it'd be for a general-purpose car?
Changing the page granularity just means you need less page table entries in your OS.
.. lots of PTEs.
Eg, if you have 4gb of ram in your machine and you're using it all, calculate how many PTEs there are.
Now, say you're running Oracle which, IIRC (and I haven't ever run Oracle, but I've done similar tricks with other bits of software) mmap()s a big ass buffer between multiple processes, if you map 1gb of RAM thats 262,144 PTEs at 4k pages. Over, say, 30 processes thats
If you flip the size bit and get the larger pages you just reduce the PTE size. The main factor here is that the external address stuff is _still_ 32 bit.
Now, PAE lets you do some bank swapping magic _WHICH_ can be useful for things like disk buffers. But for running code and mmap()ed buffers things look tricky.
2c, correct me, etc.
Marriage? Sex?
You have a file (/etc/rc.conf) which specifies configuration parameters for the various _system_ programs.
/usr/local/etc/rc.d/ in a vaguely sysv-init-script fashion (start/stop).
/etc/init.d/sendmail and its vairous links under Solaris after installing a patchset..
/var/config/ (symlink to /etc/config? My IRIX box isn't reachable from here!). So /var/config/${programname} can be accessed from chkconfig ${programname}. So, you just write all your /etc/init.d/ scripts to run chkconfig to return whether a program is meant to run or not. You don't ever have to fiddle with the symlinks in /etc/rc?.d/ to change whether stuff starts or not.
:)
Under FreeBSD, ports are started out of
Personally I find the IRIX chkconfig system quite nice. The problem with BSD is that it doesn't extend to any addon software. The problem with normal SYSV is that you rename/modify the scripts - which are then un-renamed or replaced when it comes time to upgrade. I've suffered many hours of frustration at
(For those who don't know about chkconfig - it maintains a configuration fileset in
Brilliant!
How much juice are you generating/using from the solar panels? I'm quite curious.
Do you drive a car?
Do you use a computer?
Do you, shock, grow your own food? Or buy it from the supermarket?
There's much more you can do for the environment and general well-being of the planet than worrying about the cruelty and inefficincy of animal products.
(I tried to resist, I did...
</troll>
Because, at least in Australia, huge swaths of land aren't really suitable for anything besides sheep and cattle stations.
(I forget the exact figures, but a few head per acre might be about right. Someone can flame me with the real figures.)
(Veganism as a religion aside, I don't wanna stray too far off topic..)
Search the RFC/IETF sites for URs or Universal Resource Names.
.. transparent. But then this requires caching servers everywhere and thats a thread I don't really want to get into right now.)
;-)
Squid implemented some basic URN stuff a while back - It'd run a cgi script at a URL which returned a list of sites in a specific format. Squid would then ping the sites and return them to you w/ the RTTs. You could then choose the 'best one'.
Not very effective - and I don't think it (the squid side) ever progressed past the initial stages - but maybe its the beginning for someone to tackle this issue again.
(IMHO it could be made more
Note - there's RFCs covering URNs for various things (IETF documents, ISBNs, etc)..
No, DNS isn't really designed for this. Yes, DNS has been perverted enough, so I guess this'd just be another extension. Write an RFC, get it published..