The vast majority of this space is taken up by revision histories (and those are compressed!) Periodic database dumps are available for download. Image and multimedia uploads have been taking up a bigger share lately, but those are on a separate server which recovered just fine.
A German company has published an end-user-friendly CD-ROM of material from the German-language Wikipedia, but afaik no one's published an English-language edition yet.
The colocation facility has diesel generators to protect against the outside power going out. Thanks to the miracle of circuit breakers, power circuits inside the facility shut off (including both circuits feeding our dual-power supply machines).
There's no point. Spam doesn't stay in wikipedia long enough for it to be indexed very often.
of course other wikis using mediawiki might be helped.
Right, that is the point; getting the markers in the software and out there in the wild.
The best thing i think is using it to block 'edit this page' links, and 'histories'. Even though these have noindex anyway, they still get spidered heaps.
That's a good idea, actually. On Wikipedia & co we use robots.txt to protect against that, but that's not convenient for everyone. It's somewhat unclear however whether Google actually _doesn't follow_ the links or just doesn't count them for pagerank from the documentation so far.
At this time support is limited and experimental. There is not presently a way to mark external links in a way which would cause them to be ranked, but hypothetically this could change.
Remember this thing's all of eight hours old...;)
Wikipedia isn't a link farm, though; the priority is on internal text contents and links between articles in the encyclopedia.
There is an old experimental plugin that can wrap certain types of Photoshop plugins (Win32-only). Google up "gimp photoshop plugins" and you should find it.
Also there is some support for Photoshop plugins built with the "filter factory", which apparently aren't actually native code DLLs (like most Photoshop plugins). Google will help you with that one as well.
What the article seems to be trying to say is that this protocol works better than TCP/IP does on a heavily-used connection with bandwidth at the level of 6000 times greater than a typical DSL line.
Nothing to see here, move along... it won't get grits to your home any faster.
I gave a whirl at installing Win98 Second Edition into qemu 0.5.2 (x86 binary) on my Linux box (Fedora Core 1 on an old Pentium II).
Create a hard disk image:
dd if=/dev/zero of=win98.img bs=1M count=1024
Stick in your win98 install CD and go!
qemu -cdrom/dev/cdrom -hda win98.img -boot d
At the boot menu select boot to DOS option. Run FDISK and create a primary DOS partition. Exit qemu.
Start up qemu again, this time go into Windows setup. Should be fairly standard.
At some point it may give an odd error message or two. For me it complains about being unable to allocate memory for the device manager. A bit later it said it couldn't load explorer.exe and that I'd have to reinstall Windows; just rebooting (exit, restart qemu) got it going again.
At some point after one of the reboots it'll try to install some networking stuff. For some reason I can't get Win98 to access the CD-ROM, so it can't install this and won't boot up in non-safe mode. You may want to perform this next step earlier:
Boot off the CD into DOS w/ CD-ROM support. Copy the *.CAB files from d:\win98 into c:\windows\system\precopy. Reboot and if necessary go into safe mode and fiddle with the networking control panel to get it to finish installing things.
Voila! It sort of works.
Some caveats; I haven't been at it long but here's my problems so far:
Video is VGA 16-color only, and on every boot it wants me to look for a better video driver.
Can't access CD-ROM from windows.
At least on an old Pentium II, the animated menus are _really_ slow. Turn them off!
I haven't gotten networking working yet.
Hardware detection wizard crashes reliably.
Sometimes the keyboard & mouse get locked up.
However it gets that far, which is impressive.:) And while the performance isn't super in absolute terms, it runs much faster on an old Pentium II than I ever got Bochs to run on my 2 GHz Athlon. I'm expecting good things in the future...
Why doesn't Bochs copy the usabillity of Virtual PC
Bochs is really a debugging tool for people writing their own OS. It's written to be accurate and portable, not fast or convenient. For those of us not writing our own operating systems, we're just not the target audience.
I've already extolled the virtues of QEMU's interesting capabilities and much greater speed. It's also I think a little easier to use than Bochs. It's not point and click, but it's a little more UNIX-friendly: you can run it from the command line in a sane manner compared with trying to cobble together a cryptic configuration file for Bochs.
QEMU isn't perfect, though. While the latest release will run Windows 98, it may spontaneously crash during installation, etc, and so far only runs under Linux (though a Darwin port is in the works).
QEMU has some experimental support for emulating a PowerPC (or SPARC or ARM or x86) processor, though of course it's less likely that many people would want to do so.
QEMU's not as mature as Bochs, but it's much faster, based on dynamic translation; you might think of it as a little more like a JIT compiler than an emulator. The other really interesting thing about QEMU is that in addition to a full-machine emulation mode, it can run Linux binaries from one architecture directly, translating the system call parameters as necessary. In theory at least you should be able to run binary-only x86 software -- or win32 programs on Wine -- on Linux-PPC for instance.
We were able to revive the server with disk problems and move everything onto another machine today, and Wikipedia is back online.
Again, thanks to everyone who's pitched in to support this project; nobody likes downtime, particularly not on major sites with the popularity of a Slashdot or Wikipedia. Thanks to the generous donations of many Wikipedians and Slashdotters, the next hardware failure that strikes from the blue shouldn't bring us down for two days!
In less than a day over $13000 has been donated on top what's already in the bank from past donations; we may well have surpassed the $20000 target by morning.
On behalf of all the folks at Wikipedia, a warm thanks to everyone who's contributed to the project, whether through your labor or your pocketbook!
Wikipedia doesn't run ads, so there's no other revenue stream than donations at present.
Most of the server admin and software development is done by unpaid volunteers, which is no secret.
Jimmy Wales (the founder) donates the bandwidth, the hosting space, and the time of one of his employees for hardware installation, but the new servers are additional cost that's coming from the third-party donations to the foundation.
If he were to just go kayaking with the money and leave us serverless, well you'd hear about it.;) Wikipedia is under the GNU Free Documentation License, and were there a real reason for it the community could fork the project, taking the content with them and outdoing the original site.
The purpose of the Opteron box is to throw as much memory at the database as is humanly possible, to speed up aggregate queries that check over the entire set of articles. It's got 4 gigabytes of RAM now, and is intended be upgraded in the future. With a 32-bit OS you can only put so much of that into one process space.
The databases altogether come to about 35 gigabytes (including indexes and the complete revision history of several hundred thousand wiki pages), though with some judicious compression of page text that could be brought down.
With $20K, I could handle... hmm... at least 1.2m hits a day for a year.
This traffic graph is slightly out of date (stored on the backup server) and only covers the English-language encyclopedia site, which is about half of the total traffic. (The German and Japanese-language encyclopedias are rather well trafficked as well, for instance.) It records about 850k page hits per day.
The ranking number is a rolling average; if you look at the comparitive traffic graph, you'll see that/. has held fairly steady all year while Wikipedia's been on an upward trend, and the two have recently met.
Can't say for sure if it'll even out soon or keep growing.
GNU FDL, yes, but without invariant sections. No need to attach the Wikipedia Manifesto to every article.;)
I know it's not perfect, but there was no Creative Commons when Wikipedia was started, and it's not clear what would have been better at the time.
Re:Then why is this posted to the front page?
on
Wikipedia Needs $20K
·
· Score: 4, Insightful
Your skepticism is appreciated, but inventing purchasing decisions that have not yet been made just to get your five dollars doesn't make a lot of sense.
If you don't want to donate, please don't feel any pressure to do so.
If you would like to help, please feel free to roll up your sleeves and get involved like any other volunteer and make some solid, experience-based recommendations.
Slashdot Mirror
The vast majority of this space is taken up by revision histories (and those are compressed!) Periodic database dumps are available for download. Image and multimedia uploads have been taking up a bigger share lately, but those are on a separate server which recovered just fine.
A German company has published an end-user-friendly CD-ROM of material from the German-language Wikipedia, but afaik no one's published an English-language edition yet.
The colocation facility has diesel generators to protect against the outside power going out. Thanks to the miracle of circuit breakers, power circuits inside the facility shut off (including both circuits feeding our dual-power supply machines).
Our database masters do have dual power supplies. The circuit breakers were tripped on both sides.
There's no point. Spam doesn't stay in wikipedia long enough for it to be indexed very often. of course other wikis using mediawiki might be helped.
Right, that is the point; getting the markers in the software and out there in the wild.
The best thing i think is using it to block 'edit this page' links, and 'histories'. Even though these have noindex anyway, they still get spidered heaps.
That's a good idea, actually. On Wikipedia & co we use robots.txt to protect against that, but that's not convenient for everyone. It's somewhat unclear however whether Google actually _doesn't follow_ the links or just doesn't count them for pagerank from the documentation so far.
At this time support is limited and experimental. There is not presently a way to mark external links in a way which would cause them to be ranked, but hypothetically this could change.
Remember this thing's all of eight hours old... ;)
Wikipedia isn't a link farm, though; the priority is on internal text contents and links between articles in the encyclopedia.
There is an old experimental plugin that can wrap certain types of Photoshop plugins (Win32-only). Google up "gimp photoshop plugins" and you should find it. Also there is some support for Photoshop plugins built with the "filter factory", which apparently aren't actually native code DLLs (like most Photoshop plugins). Google will help you with that one as well.
Wouldn't the LGPL allow them to link with a binary-only copy protection module?
Not at the moment that I know of, but MoL is apparently being ported to Mac OS X. (This comes up on the mailing list from time to time.)
What the article seems to be trying to say is that this protocol works better than TCP/IP does on a heavily-used connection with bandwidth at the level of 6000 times greater than a typical DSL line.
Nothing to see here, move along... it won't get grits to your home any faster.
Admittedly, a 12 MHz PC was pretty good in 1988...
Yes, qemu emulates an NE2000 which can connect to the host via the tun interface.
- Create a hard disk image:
- Stick in your win98 install CD and go!
/dev/cdrom -hda win98.img -boot d
- At the boot menu select boot to DOS option. Run FDISK and create a primary DOS partition. Exit qemu.
- Start up qemu again, this time go into Windows setup. Should be fairly standard.
- At some point it may give an odd error message or two. For me it complains about being unable to allocate memory for the device manager. A bit later it said it couldn't load explorer.exe and that I'd have to reinstall Windows; just rebooting (exit, restart qemu) got it going again.
- At some point after one of the reboots it'll try to install some networking stuff. For some reason I can't get Win98 to access the CD-ROM, so it can't install this and won't boot up in non-safe mode. You may want to perform this next step earlier:
- Boot off the CD into DOS w/ CD-ROM support. Copy the *.CAB files from d:\win98 into c:\windows\system\precopy. Reboot and if necessary go into safe mode and fiddle with the networking control panel to get it to finish installing things.
- Voila! It sort of works.
Some caveats; I haven't been at it long but here's my problems so far:dd if=/dev/zero of=win98.img bs=1M count=1024
qemu -cdrom
However it gets that far, which is impressive. :) And while the performance isn't super in absolute terms, it runs much faster on an old Pentium II than I ever got Bochs to run on my 2 GHz Athlon. I'm expecting good things in the future...
Of course either way it's speculation at this point.
Bochs is really a debugging tool for people writing their own OS. It's written to be accurate and portable, not fast or convenient. For those of us not writing our own operating systems, we're just not the target audience.
I've already extolled the virtues of QEMU's interesting capabilities and much greater speed. It's also I think a little easier to use than Bochs. It's not point and click, but it's a little more UNIX-friendly: you can run it from the command line in a sane manner compared with trying to cobble together a cryptic configuration file for Bochs.
QEMU isn't perfect, though. While the latest release will run Windows 98, it may spontaneously crash during installation, etc, and so far only runs under Linux (though a Darwin port is in the works).
QEMU's not as mature as Bochs, but it's much faster, based on dynamic translation; you might think of it as a little more like a JIT compiler than an emulator. The other really interesting thing about QEMU is that in addition to a full-machine emulation mode, it can run Linux binaries from one architecture directly, translating the system call parameters as necessary. In theory at least you should be able to run binary-only x86 software -- or win32 programs on Wine -- on Linux-PPC for instance.
Again, thanks to everyone who's pitched in to support this project; nobody likes downtime, particularly not on major sites with the popularity of a Slashdot or Wikipedia. Thanks to the generous donations of many Wikipedians and Slashdotters, the next hardware failure that strikes from the blue shouldn't bring us down for two days!
On behalf of all the folks at Wikipedia, a warm thanks to everyone who's contributed to the project, whether through your labor or your pocketbook!
Jimmy Wales (the founder) donates the bandwidth, the hosting space, and the time of one of his employees for hardware installation, but the new servers are additional cost that's coming from the third-party donations to the foundation.
If he were to just go kayaking with the money and leave us serverless, well you'd hear about it. ;) Wikipedia is under the GNU Free Documentation License, and were there a real reason for it the community could fork the project, taking the content with them and outdoing the original site.
See MeatBall:RightToFork.
The databases altogether come to about 35 gigabytes (including indexes and the complete revision history of several hundred thousand wiki pages), though with some judicious compression of page text that could be brought down.
It's also cool, of course. ;)
Our offsite backup MX is running just fine pumping messages to their destinations, but subscription may not work properly.
This traffic graph is slightly out of date (stored on the backup server) and only covers the English-language encyclopedia site, which is about half of the total traffic. (The German and Japanese-language encyclopedias are rather well trafficked as well, for instance.) It records about 850k page hits per day.
Can't say for sure if it'll even out soon or keep growing.
I know it's not perfect, but there was no Creative Commons when Wikipedia was started, and it's not clear what would have been better at the time.
If you don't want to donate, please don't feel any pressure to do so.
If you would like to help, please feel free to roll up your sleeves and get involved like any other volunteer and make some solid, experience-based recommendations.
I don't have up-to-the-minute numbers, but several hours ago it was up about $2000 from the initial figure.