I expect you can get an Intel 1000/Pro for around $30; full TCP/IP checksum offloads in both directions, interrupt moderation, jumbo frames, and Intel even write their own open source drivers.
Heh, my on-board Realtek GigE chip has checksum offloads too, but even with them on, 300Mbps would have me up to 70% system/interrupt CPU load (and I hear the checksumming is a bit.. broken); I barely scrape 30% with a PRO/1000GT.
Slightly flawed in that it uses a constant salt, meaning once a set of rainbow tables have been generated all your passwords are easily compromised.
I had a patch to it that supported non-salted, statically-salted and dynamically salted passwords (upgrading on auth) without changing table schemas beyond the default 38 bytes. I don't have it to hand unfortunately, but it's not hard to do; just use pack() to b64 the hash, put in some random bytes next to it seperated from the hash by $ or so, attempt each older method on failure, and now those rainbow tables are only good for a single password.
"Ruby's lack of opcode cache concerns me regarding performance."
With FastCGI you've just got a Ruby daemon just like any other; it's entirely persistant during the lifetime of the server -- you don't leave the interpreter and reset all that state at the end of each request. This is pretty important with all the setup Rails does with dependency injection, Routes etc.
mod_ruby behaves similarly AIUI; it's a single persistant copy of the interpreter, rather than starting from scratch each time; opcode caching is simply not relevent in these situations since the interpreter is always "hot".
TFT's are either sold as having 16.2 million colours, or 16.7 -- the 16.2m ones have 6 bits per subpixel and use temporal dithering (rapidly flickering subpixels between two colours) to make their quarter of a million native colours stretch to the 16.2m perceived ones; the 16.7m ones have 8 bits per subpixel, and display the full range natively.
"RAID 0+1 mirrors and spans, but a 4 disk RAID-5 array has 50% more space than a 4 disk RAID 0+1"
On the other hand, if a disk fails on your RAID-5 array, you're liable to see a significant drop in performance, since it starts having to rebuild data from parity spread across all the disks. Have a disk fail in your RAID-10 array and the performance drop is going to be much less significant since it's just a single degraded RAID-1. The RAID-10* array also has a significantly better chance of surviving if two disks die.
This is critical if you're going for speed and redundancy (say, on a busy database server) over redundancy and space (like your network file server).
* That is, 2 mirrors, and a stripe across them; the other way around is Very Bad[tm].
Generally, no; especially in the case of cryptographic hash algorithms like SHA*, MD5, Whirlpool and friends, which are designed so that even very minor changes in their inputs result in (with overwhelming probability) completely different outputs.
However, there are algorithms for generating hashes which change only slightly with similar inputs; Nilsimsa is one such example, used by a number of spam detection systems. What you suggest would seem quite feasible using something like this, though they wouldn't be a replacement for more secure algorithms for determining file integrity.
You know, part of me thinks this kind of thing would actually be kinda cool.. but not from an advertising perspective. I'm thinking more along the lines of a recognisable object in space; maybe some kind of slowly spinning inflated shape, with markings delimiting its size, coloured fairly dim so it's visible but not about to drown out the stars.
Think of all the people you'd make look at the night sky, get out their binoculars and telescopes, look at astronomy magazines and websites to track it, find out about the technologies involved, not to mention see an undeniable physical reference showing the sort of scales involved.
I keep hearing this from.. certain directions, but I'm yet to see an actual explanation as to why GEOM sucks more than the ad-hoc mess it replaces. GEOM's given us better RAID support, the ability to export block devices across networks, disk encryption, better support for multiple partition formats and disk layouts, and a rational layered approach which allows for pretty much arbitary nesting of any of the above to suit whatever you want to do, not to mention a nice, well documented API for developing your own classes. Frankly I was kind of shocked to find this wasn't already the case.
So come on Mr Uid Half A. Million Eighty Six Thousand And Very Odd, earn your 30% Insightful moderation and enlighten us all.
Quite. I receive Sky One, but I'm useless at remembering when things are on and actually being at the TV downstairs which is the only one in the house which has a STB. So.. I downloaded every single episode of BSG.
Last week, I noticed the DVD box set was available, and bought it, along with the mini-series. I'm not much for TV, but this is the fourth TV series I've bought after downloading episodes; I've bought a lot more movies following the same pattern. Ditto for CD's (although I make an exception for "protected" CD's; if I can't rip it, it's useless to me, and I have no desire to support them), games, apps, and even a few books.
Prior to easy downloading of these things, I rarely bought anything; TV is inconvenient, rental places seem mainly to stock junk (and where's my IMDB terminal?), and I don't listen to radio or keep up with review sites. The only people who suffer are those who would sell me things I dislike, and why the hell should I support *them*? Isn't capitalism supposed to work better when everyone has perfect knowledge?
Capacity isn't the only metric of a drive; SCSI drives generally concentrate on keeping seek times low, and they do this by using lower areal density (so the heads don't need to spend so long settling on a track) and physically smaller platters (so they have less distance to travel), amongst other things. I dare say this also contributes to their robustness; lower density = lower error rates, and a greater ability to cope with minor defects.
Yes, turck-mmcache (now eAccelerator) works fine with PHP/FastCGI; we make heavy use of it:)
Re:But does it pass the ACID2 Test?
on
Opera 8 Released
·
· Score: 2, Informative
The invalid CSS is deliberate; browsers are expected to skip the broken sections, since that's how CSS handles forward compatibility (i.e. a CSS 2 client can handle CSS 3 by skipping the bits it can't parse).
It's supposed to trip up browsers that don't follow the spec by giving them invalid but potentially parsable rules which they shouldn't apply.
Per CPU, according to one (admittedly somewhat vague) document from AMD. Also an old example system with 8 slots per processor. And here's a more recent motherboard with an 8+4 configuration: K8D Master 3.
Any noticable speed loss is unlikely; that you need fewer PHP processes and database connections eating memory could actually make it faster.
There's no need for #! lines in your.php files; the interpreter loads whatever SCRIPT_FILENAME is set to by the webserver. With external servers you don't even need the.php's to be in the webserver's document tree; your PHP interpreter(s) can be running in a seperate part of the filesystem, a chroot jail or even a seperate server; even load balanced across many servers using something like the pen load balancer or lighttpd and its native lb support.
Not necessarily the simplest or most sensible setup for small sites/development, but it works well in more loaded setups.
PHP itself is thread safe (that's what the Thread Safe Resource Manager is all about); if your extension use is limited, it should work fine in a threaded MPM.
A better solution IMO is to use FastCGI; mod_fastcgi is thread safe, and it seperates PHP (and pretty much any other language) from the webserver so you can't make the webserver blow up with a badly written script, as well as using fewer resources since there's no need for a 1:1 mapping between webserver processes and PHP processes. You can even host your PHP on a seperate machine, or a cluster of them, without having to manage webservers on them all.
"I work at the local computer repair shop while going to school, and right now we check every incoming system for bristling capacitors. About 25% of the time they have bad capacitors. Why? Heat from the CPU is causing them to overheat, expand, and become useless."
"+ Focus follows mouse: no civilized human should be without this."
Seriously? I tried it for a few days recently and found it maddening that every time I so much as looked at the mouse my focus shifted. It's bad enough when certain apps/dialogs steal focus without it changing randomly because I knocked the mouse a couple of mm.
It's also utterly useless when using Synergy for two keyboards/single mouse control, since your focus shifts every time you move the cursor between machines.
"+ A file system that can distinguish between uppercase and lowercase letters."
Um? I'm pretty sure every filesystem I've used in the past decade or so has been at least case-preserving. Actual case sensitivity is more down to taste.
Not really relevent as far as I'm concerned. I use FLAC so I'm not stuck with format-of-the-month from 3 years ago, and to make transcoding something I can do without feeling sick, and to utterly obliviate any worries about edge cases where my favourite lossy codec falls short (they all do at some point.. what are you going to do, ABX every single track you'll ever listen to?).
Disk space is cheap and plentiful, and is only getting more so; why should it bother me if my music takes a few extra percent of my storage?
"let's say that unbroken SHA-1 represents a 100 meter (328 ft) wall. if a
break allows a collision to be found in merely 2^69 operations (on average), that would mean the wall has crumbled to 4.9 cm (1.9 in) tall. that's broken!!
OTOH, let's say that unbroken MD5 represents a 100 meter (328 ft) wall. comparing unbroken MD5 to broken SHA-1 means the wall would actually grow from 100 meters (328 ft) tall to 3.2 km (1.99 miles) tall. SHA-1, even if it's broken enough to find a collision in 2^69 operations (on average), is still stronger than MD5 was ever meant to be.
again, using unbroken MD5 as our reference of a 100 meter (328 ft) wall, unbroken SHA-1 would be a wall 6553.6 km (4072 miles) tall. SHA-1 was intended to be incredibly stronger than MD5."
DRM which prevents me from playing music in my chosen music player is not exactly something I think I'm going to fail to notice. If you want to supply me with music, video, images, or text you do so using an open standard I can use in my choice of system, sans arbitary, pointless and impossible to enforce technical restrictions, or you supply me with nothing at all.
Say "but you can burn it to CD and rip that" and I *will* hurt you.
"we'd better get used to it."
*shrug*, get used to it if you like. Personally I'll just stick with those willing to supply what I demand. Something I can only use in a single crappy media player is not it, I don't care how lax you make the restrictions in it, especially if you get to change them arbitarily after the fact.
Media companies are in a perfect position to undercut pirates; I can get rips of pretty much any music or movie in practically any format I desire with varying amounts of effort and success. If I could just go and buy a FLAC, or set of VOBs or an AVI from www.some-publisher.com and get consistant download speeds and quality, well organized selections and the formats I want, why the hell would I go hunting for them on $crappy_pirate_network_of_the_week?
Those who recognise this will get my custom. The rest I can live without.
Slashdot Mirror
I expect you can get an Intel 1000/Pro for around $30; full TCP/IP checksum offloads in both directions, interrupt moderation, jumbo frames, and Intel even write their own open source drivers.
Heh, my on-board Realtek GigE chip has checksum offloads too, but even with them on, 300Mbps would have me up to 70% system/interrupt CPU load (and I hear the checksumming is a bit.. broken); I barely scrape 30% with a PRO/1000GT.
"Salted Hash Login Generator"
Slightly flawed in that it uses a constant salt, meaning once a set of rainbow tables have been generated all your passwords are easily compromised.
I had a patch to it that supported non-salted, statically-salted and dynamically salted passwords (upgrading on auth) without changing table schemas beyond the default 38 bytes. I don't have it to hand unfortunately, but it's not hard to do; just use pack() to b64 the hash, put in some random bytes next to it seperated from the hash by $ or so, attempt each older method on failure, and now those rainbow tables are only good for a single password.
"Ruby's lack of opcode cache concerns me regarding performance."
With FastCGI you've just got a Ruby daemon just like any other; it's entirely persistant during the lifetime of the server -- you don't leave the interpreter and reset all that state at the end of each request. This is pretty important with all the setup Rails does with dependency injection, Routes etc.
mod_ruby behaves similarly AIUI; it's a single persistant copy of the interpreter, rather than starting from scratch each time; opcode caching is simply not relevent in these situations since the interpreter is always "hot".
TFT's are either sold as having 16.2 million colours, or 16.7 -- the 16.2m ones have 6 bits per subpixel and use temporal dithering (rapidly flickering subpixels between two colours) to make their quarter of a million native colours stretch to the 16.2m perceived ones; the 16.7m ones have 8 bits per subpixel, and display the full range natively.
'"Wow, that would have taken me about 3 days to code that in PHP."'
Funny, that's pretty much what PHP people moving to Ruby/Rails keep saying too. Not that it's exactly difficult to beat PHP, but still..
"RAID 0+1 mirrors and spans, but a 4 disk RAID-5 array has 50% more space than a 4 disk RAID 0+1"
On the other hand, if a disk fails on your RAID-5 array, you're liable to see a significant drop in performance, since it starts having to rebuild data from parity spread across all the disks. Have a disk fail in your RAID-10 array and the performance drop is going to be much less significant since it's just a single degraded RAID-1. The RAID-10* array also has a significantly better chance of surviving if two disks die.
This is critical if you're going for speed and redundancy (say, on a busy database server) over redundancy and space (like your network file server).
* That is, 2 mirrors, and a stripe across them; the other way around is Very Bad[tm].
Generally, no; especially in the case of cryptographic hash algorithms like SHA*, MD5, Whirlpool and friends, which are designed so that even very minor changes in their inputs result in (with overwhelming probability) completely different outputs.
However, there are algorithms for generating hashes which change only slightly with similar inputs; Nilsimsa is one such example, used by a number of spam detection systems. What you suggest would seem quite feasible using something like this, though they wouldn't be a replacement for more secure algorithms for determining file integrity.
You know, part of me thinks this kind of thing would actually be kinda cool.. but not from an advertising perspective. I'm thinking more along the lines of a recognisable object in space; maybe some kind of slowly spinning inflated shape, with markings delimiting its size, coloured fairly dim so it's visible but not about to drown out the stars.
Think of all the people you'd make look at the night sky, get out their binoculars and telescopes, look at astronomy magazines and websites to track it, find out about the technologies involved, not to mention see an undeniable physical reference showing the sort of scales involved.
"geom sucks anyhow"
I keep hearing this from.. certain directions, but I'm yet to see an actual explanation as to why GEOM sucks more than the ad-hoc mess it replaces. GEOM's given us better RAID support, the ability to export block devices across networks, disk encryption, better support for multiple partition formats and disk layouts, and a rational layered approach which allows for pretty much arbitary nesting of any of the above to suit whatever you want to do, not to mention a nice, well documented API for developing your own classes. Frankly I was kind of shocked to find this wasn't already the case.
So come on Mr Uid Half A. Million Eighty Six Thousand And Very Odd, earn your 30% Insightful moderation and enlighten us all.
Quite. I receive Sky One, but I'm useless at remembering when things are on and actually being at the TV downstairs which is the only one in the house which has a STB. So.. I downloaded every single episode of BSG.
Last week, I noticed the DVD box set was available, and bought it, along with the mini-series. I'm not much for TV, but this is the fourth TV series I've bought after downloading episodes; I've bought a lot more movies following the same pattern. Ditto for CD's (although I make an exception for "protected" CD's; if I can't rip it, it's useless to me, and I have no desire to support them), games, apps, and even a few books.
Prior to easy downloading of these things, I rarely bought anything; TV is inconvenient, rental places seem mainly to stock junk (and where's my IMDB terminal?), and I don't listen to radio or keep up with review sites. The only people who suffer are those who would sell me things I dislike, and why the hell should I support *them*? Isn't capitalism supposed to work better when everyone has perfect knowledge?
Capacity isn't the only metric of a drive; SCSI drives generally concentrate on keeping seek times low, and they do this by using lower areal density (so the heads don't need to spend so long settling on a track) and physically smaller platters (so they have less distance to travel), amongst other things. I dare say this also contributes to their robustness; lower density = lower error rates, and a greater ability to cope with minor defects.
This is to test browser CSS forward-compatible error parsing rules. If a browser fails to skip these lines it fails the test.
Yup; it's even one of the three great virtues of a programmer. Gotta be the right kind of laziness though ;)
Yes, turck-mmcache (now eAccelerator) works fine with PHP/FastCGI; we make heavy use of it :)
The invalid CSS is deliberate; browsers are expected to skip the broken sections, since that's how CSS handles forward compatibility (i.e. a CSS 2 client can handle CSS 3 by skipping the bits it can't parse).
It's supposed to trip up browsers that don't follow the spec by giving them invalid but potentially parsable rules which they shouldn't apply.
Per CPU, according to one (admittedly somewhat vague) document from AMD. Also an old example system with 8 slots per processor. And here's a more recent motherboard with an 8+4 configuration: K8D Master 3.
Any noticable speed loss is unlikely; that you need fewer PHP processes and database connections eating memory could actually make it faster.
.php files; the interpreter loads whatever SCRIPT_FILENAME is set to by the webserver. With external servers you don't even need the .php's to be in the webserver's document tree; your PHP interpreter(s) can be running in a seperate part of the filesystem, a chroot jail or even a seperate server; even load balanced across many servers using something like the pen load balancer or lighttpd and its native lb support.
There's no need for #! lines in your
Not necessarily the simplest or most sensible setup for small sites/development, but it works well in more loaded setups.
AMD64's support 8 slots per CPU; that just happens to be a bit much to fit on most motherboards.
PHP itself is thread safe (that's what the Thread Safe Resource Manager is all about); if your extension use is limited, it should work fine in a threaded MPM.
A better solution IMO is to use FastCGI; mod_fastcgi is thread safe, and it seperates PHP (and pretty much any other language) from the webserver so you can't make the webserver blow up with a badly written script, as well as using fewer resources since there's no need for a 1:1 mapping between webserver processes and PHP processes. You can even host your PHP on a seperate machine, or a cluster of them, without having to manage webservers on them all.
"I work at the local computer repair shop while going to school, and right now we check every incoming system for bristling capacitors. About 25% of the time they have bad capacitors. Why? Heat from the CPU is causing them to overheat, expand, and become useless."
Are you sure it's not this well known capacitor problem?
The codec portions (libflac and friends) are BSD licensed; the FLAC tools themselves are GPL.
WavPack is another nice lossless BSD-licensed codec, which is more advanced in some respects, if not as well supported.
"+ Focus follows mouse: no civilized human should be without this."
Seriously? I tried it for a few days recently and found it maddening that every time I so much as looked at the mouse my focus shifted. It's bad enough when certain apps/dialogs steal focus without it changing randomly because I knocked the mouse a couple of mm.
It's also utterly useless when using Synergy for two keyboards/single mouse control, since your focus shifts every time you move the cursor between machines.
"+ A file system that can distinguish between uppercase and lowercase letters."
Um? I'm pretty sure every filesystem I've used in the past decade or so has been at least case-preserving. Actual case sensitivity is more down to taste.
Not really relevent as far as I'm concerned. I use FLAC so I'm not stuck with format-of-the-month from 3 years ago, and to make transcoding something I can do without feeling sick, and to utterly obliviate any worries about edge cases where my favourite lossy codec falls short (they all do at some point.. what are you going to do, ABX every single track you'll ever listen to?).
Disk space is cheap and plentiful, and is only getting more so; why should it bother me if my music takes a few extra percent of my storage?
DRM which prevents me from playing music in my chosen music player is not exactly something I think I'm going to fail to notice. If you want to supply me with music, video, images, or text you do so using an open standard I can use in my choice of system, sans arbitary, pointless and impossible to enforce technical restrictions, or you supply me with nothing at all.
Say "but you can burn it to CD and rip that" and I *will* hurt you.
"we'd better get used to it."
*shrug*, get used to it if you like. Personally I'll just stick with those willing to supply what I demand. Something I can only use in a single crappy media player is not it, I don't care how lax you make the restrictions in it, especially if you get to change them arbitarily after the fact.
Media companies are in a perfect position to undercut pirates; I can get rips of pretty much any music or movie in practically any format I desire with varying amounts of effort and success. If I could just go and buy a FLAC, or set of VOBs or an AVI from www.some-publisher.com and get consistant download speeds and quality, well organized selections and the formats I want, why the hell would I go hunting for them on $crappy_pirate_network_of_the_week?
Those who recognise this will get my custom. The rest I can live without.