Debian Lenny (which was released last weekend or so) supports the Eee out of the box. (I'm using it on an Eee1000 now. http://wiki.debian.org/DebianEeePC has links to a custom installer which will install all the right packages for you, but it's juts Debian under the hood.
I get good battery life and everything seems to pretty much Just Work(tm)
my primary server right now has a 4GB swap partition and 1.25GB of RAM... a piece of bad AJAX code that ran overnight wound up using all the RAM and had some seriously detrimental effects on the performance of the server. it took 25 minutes to ssh in in the morning and when I finally got in, I found that the load averages were at over 100 (I've NEVER see that before).
OTOH, would it not have been better to let memory fill up and the OOM killer kill your AJAX code? Rather than stopping your server from responding overnight. I've seen machines requiring console access or a forced reboot to recover from that.
We have several servers where the reduced responsiveness of hitting swap is as bad as them going away, we'd rather make sure that the only time it can run out of memory is with a rogue process and hope the OOM killer reaps it first.
In the UK Andrews and Arnold (www.aaisp.net.uk) will do you natively routed IPv6 or IPv6 over a 6to4 tunnel if your modem/router doesn't support native IPv6.
/etc/init.d/ scripts: Debian doesn't have a nice way to turn these scripts on and off and monitor their status via a command-line tool. Red Hat's system here was very good.
update-rc.d
package management: Yeah, apt-get's dependency resolution logic is very cool. Other aspects of the system aren't so cool. Apt-get, aptitude, and other front-ends don't share the same back-end data-store, so if you mix and match these tools, you get inconsistent package data. And it's nearly impossible to force-remove a package (just delete all the damn files and forget about it!) if the associated removal script fails.
This isn't true, all these packages use the dpkg package list, and install packages with dpkg. The package list will always be in sync between them. Some of the fringe things *are* different (dpkg --set-selections holding, for example), but I consider this a bug.
If you are going to install cutting edge ports from cvs on freebsd, you should compare it to installing packages from Debian Unstable, not the stable branch. If you are willing to put up with possible system-breaking changes from newer versions being installed, then Testing and Unstable (currently Etch and Sid) are there for you. If you want a guarantee that once setup your system will stay that way, you want Debian Stable.
I agree with this entirely. Metaframe is designed exactly for the situation where you want to distribute one application to a large number of clients, whether they be running windows or linux.
With the single application mode you can run a single application like you would using X forwarding and there are loads of really cool optimisations they have done recently.
I'd recommend using the Java client rather than the native linux one, however. Eww Motif.
ROX (RiscOS On X) which has a filer, window manager and a session manager uses Application Directories taken from Risc OS. This sounds very similar to Apple Application Bundles.
Installation is done by copying the directory, and the first time you run it, it will be compiled. You do have to run it from ROX-Filer for this to be supported (just double-click on the application directory), otherwise you have to run a script inside the directory.
Recently ROX has combined AppDirs with the Zero Install installation method, which uses a caching-remote filesystem. You can run things direct from the server they are distributed on using a virtual filesystem which will locally cache the files.
Having personally written some of the printer support for the Java client I can tell you that the printer support is actually quite good now. And doesn't involve running random drivers in kernel mode on the server,
Also, having worked in the Research Lab for a summer I happen to know that quite a lot of the issues mentioned above (text latency, graphics slowness etc) are being fixed in the next few releases. They have some *really* cool tech, which I'm probably not allowed to tell people about.
I imagine its similar to the Xen project at the lab here. The Xen hyper-visor runs in ring 0, and linux was ported to it by removing all the ring-0 calls and making them calls to Xen instead. This was fairly easy, as linux is designed for portability, so all these (hardware-dependent) calls are abstracted well.
I was talking to someone working on the XP port, who said that it really is not designed for portability any more, and that even with the source code its a lot of work to port it.
Have a look at the Dasher Project. Its not voice input, but they have several input methods which can be used by para- and quadra-plegics (eye tracker, head mouse, breath mouse, toungue mouse, etc) - you pretty much can use any movement you can make. It can be faster and more accurate than most voice recognition systems, and they have it working for controlling the UI as well as textual input now.
Seriously, though, when the copyright on a work expires, the previous holder of that copyright is under no legal, or moral, compulsion to make it easy, or even possible, for you to make copies of the work.
No, but the law should make it illegal to do it. They don't have to make it possible - but if we manage anyway we're allowed. I don't mind companies making it hard for me to do something (well, I'll have an issue with the company, and complain about *them*), but the government should be supporting *my* right to make such a copy, not supporting them at stopping me.
Actually, if you haven't tried it since
version 5 or 6, you'll find the rendering
engine is as good as mozilla, and occasionally
better. Its now a case of specific site-
variation, rather than one being obviously
better. The main re-write in this version has
been the CSS engine - which is superb now,
and renders anything I throw at it.
The main thing I like about Opera though is
the mouse gestures - now customisable.
(I know, Mozilla has them too, but I still
find Opera a lot more responsive, and nicer
to use.
This is very true. To teach a strong theoretical grounding there is little time to study multiple implementations of the theory. Instead, you concentrate on one implementation as an example of the theory.
In languages for example, you will study the concepts, but pick one (say C++) for practical work. You will look at what C++ has and doesn't have relative to the theoretical model, and (as text book study) relative to other languages such as Java and Smalltalk.
In my experience, students who have a good theoretical grounding and some practical experience with a single product can easily learn to apply the theory to other products. This is especially true for languages and common office products (word processors, spreadsheets, etc); but conversion becomes more difficult when dealing with specific applications (even IIS vs Apache configuration).
I'm currently at Cambridge University (the UK one...), and this is precisely what they do here. The course is mostly a lot higher level than specific implementations, a lot of design patterns and general algorithms, and so on. Java is taught as a language, but the aim is to use that to demonstrate principles. Its also useful for the group projects (to teach you about team work, and software development, not coding) to have a standard language.
When they do have implementation specific things, they try to put in a range. They teach ML and Java, and insist that the excercises for one are written in windows, and the other in unix, and so on. They also have comparative {OS,programming languages,etc} courses.
The aim is to produce people who aren't tied to one thing - since that will probably have changed while they are doing their undergrad, but can adapt.
Professor Sir Alec Jeffreys said the practice was highly discriminatory..... he called for the creation of a national database, storing the profiles of the entire UK population... "If we're all on the database, we're all in exactly the same boat - the issue of discrimination disappears," he said.
The discrimination might not be an issue, but its still a huge privacy violation!
I agree, particuarly with.org, but also other tlds. They should be restricted to the purpose they were created for.
.org/.plc/.ltd - require proof of status
global tlds - require some sort of evidence of being a global company (there are country codes for a reason)
and finally a tld where there are no such restrictions, but it is strictly first come first serve - if I want to register microsoft.whatever, fine - and it can't be taken back for using their name, companies have.com and.co.cc and so on.
I've got the latest version of Opera - and it doesn't complain about the certificate. It doesn't actually load the site mind you - you don't get a page that you can view the source of. Possibly also vulnerable.
Oh, and IE 5.0 did complain about the certificate being invalid
It signs messages and files just fine. What's wrong with using it?
The problem you have with this is not the signing itself, but with key trust relationships. Basically I can create a key with any name & email address - how do you know that it is actually me? There are several methods to solve this. Either you have to meet in person and give them the key there, or it has to be signed by a mutually trusted 3rd party. Meeting in person usually defies the point of public key encryption - you might as well just use symmetric if you can give them the key. The second option is a trusted 3rd party - but who do you trust that much, if you recall there was a slashdot story a while back (can't find the link) about Verisign signing dodgy Microsoft certificates.
Gnupg actually uses a web of trust system - if you have verified someone's key in person, and you have signed it to say you trust it, then it assumes you trust any key they sign. This introduces another weak point though. Even if you only sign keys after personally verifying fingerprints, do all the people who you've signed?
There are several program out there you can use, but if you are interfacing with customers using this, either you verifying their signatures, or they checking yours - there are going to be 2 problems. Do they trust the tech in general, and are they prepared to trust it, and how do you trust each other's keys. There probably needs to be a high profile 3rd party that you can both trust to authenticate against - but are there any at the moment for use with general signing tools? (Something like the certificates used in https/SSL, probably)
My dad is an airtraffic controller, and somewhat clueful about computers so I've found out what the problem here is.
Basically, there is one system that holds all the data about which flight is assigned to which transponder ID number, and what air routes etc. This is a big mainframe, currently running on 3 IBM S390s (I saw them when they were putting them in last time I was in there). The actual code, however, is the same as they were using on the original S320, and has survived 2 replacement systems, by the cunning use of emulators. Incidentally the code is so old, it doesn't store dates (and hence survived the millenium bug) - it just rolls over at midnight - all the numbers get reset / reassigned.
This code crashes fairly regularly, but until now wasn't a serious problem - because they could just restart it. Unfortunately, the new centre in Swanwick produced some new issues. The new centre takes a data feed from the mainframe at the old site. The designers of the 2 systems didn't consider the fact that the two might become out of sync, and hence when the mainframe gets reset, they go to manual in the interim, and the 2 systems are out of sync when they come back up. At one point they had to wait until midnight, reset both systems (because everything rolls over at midnight) and go from there. They appear to be able to bring the systems back in sync, and back to normal speed, within a few hours now.
I suspect the software upgrade was to fix this problem, but didn't go according to plan.
Of course the new centre itself is another story. As pointed out there were some bad design decisions (The American's - who were going to make it cheaper by doing all the devopment - they stopped, lets carry on!!) Plus the fact that Lockheed Martin were going to walk out on the contract, because the penalty clause was going to be less than the cost of finishing the project....
Slashdot Mirror
Debian Lenny (which was released last weekend or so) supports the Eee out of the box. (I'm using it on an Eee1000 now. http://wiki.debian.org/DebianEeePC has links to a custom installer which will install all the right packages for you, but it's juts Debian under the hood. I get good battery life and everything seems to pretty much Just Work(tm)
After last week and this I suggest defrauding the NYSE. Google bomb old news stories about Northern Rock and then short their stock.
Alternatively, I hear phishing scams are quite good these days. Or credit card fraud.
Matt
I found out about it from no2id (http://www.no2id.net/), sounds like you should sign up and get their bulletins
Matt
OTOH, would it not have been better to let memory fill up and the OOM killer kill your AJAX code? Rather than stopping your server from responding overnight. I've seen machines requiring console access or a forced reboot to recover from that.
We have several servers where the reduced responsiveness of hitting swap is as bad as them going away, we'd rather make sure that the only time it can run out of memory is with a rogue process and hope the OOM killer reaps it first.
In the UK Andrews and Arnold (www.aaisp.net.uk) will do you natively routed IPv6 or IPv6 over a 6to4 tunnel if your modem/router doesn't support native IPv6.
update-rc.d
This isn't true, all these packages use the dpkg package list, and install packages with dpkg. The package list will always be in sync between them. Some of the fringe things *are* different (dpkg --set-selections holding, for example), but I consider this a bug.
If you are going to install cutting edge ports from cvs on freebsd, you should compare it to installing packages from Debian Unstable, not the stable branch. If you are willing to put up with possible system-breaking changes from newer versions being installed, then Testing and Unstable (currently Etch and Sid) are there for you. If you want a guarantee that once setup your system will stay that way, you want Debian Stable.
I agree with this entirely. Metaframe is designed exactly for the situation where you want to distribute one application to a large number of clients, whether they be running windows or linux.
With the single application mode you can run a single application like you would using X forwarding and there are loads of really cool optimisations they have done recently.
I'd recommend using the Java client rather than the native linux one, however. Eww Motif.
The paper from the talk has now been released, I've mirrored it in case the site gets hit.
ROX (RiscOS On X) which has a filer, window manager and a session manager uses Application Directories taken from Risc OS. This sounds very similar to Apple Application Bundles.
Installation is done by copying the directory, and the first time you run it, it will be compiled. You do have to run it from ROX-Filer for this to be supported (just double-click on the application directory), otherwise you have to run a script inside the directory.
Recently ROX has combined AppDirs with the Zero Install installation method, which uses a caching-remote filesystem. You can run things direct from the server they are distributed on using a virtual filesystem which will locally cache the files.
There are already a lot of applications written for this.
Having personally written some of the printer support for the Java client I can tell you that the printer support is actually quite good now. And doesn't involve running random drivers in kernel mode on the server,
Also, having worked in the Research Lab for a summer I happen to know that quite a lot of the issues mentioned above (text latency, graphics slowness etc) are being fixed in the next few releases. They have some *really* cool tech, which I'm probably not allowed to tell people about.
Doesn't work on my 2.6.5 already. May be the -bk1 patch
I imagine its similar to the Xen project at the lab here. The Xen hyper-visor runs in ring 0, and linux was ported to it by removing all the ring-0 calls and making them calls to Xen instead. This was fairly easy, as linux is designed for portability, so all these (hardware-dependent) calls are abstracted well.
I was talking to someone working on the XP port, who said that it really is not designed for portability any more, and that even with the source code its a lot of work to port it.
No its not - its a spoon
Have a look at the Dasher Project. Its not voice input, but they have several input methods which can be used by para- and quadra-plegics (eye tracker, head mouse, breath mouse, toungue mouse, etc) - you pretty much can use any movement you can make. It can be faster and more accurate than most voice recognition systems, and they have it working for controlling the UI as well as textual input now.
Matthmm....and in other news thieves will now have their right hands cut off and cars parked on double yellow lines will be blown up by swat teams
Seriously, though, when the copyright on a work expires, the previous holder of that copyright is under no legal, or moral, compulsion to make it easy, or even possible, for you to make copies of the work.
No, but the law should make it illegal to do it. They don't have to make it possible - but if we manage anyway we're allowed. I don't mind companies making it hard for me to do something (well, I'll have an issue with the company, and complain about *them*), but the government should be supporting *my* right to make such a copy, not supporting them at stopping me.
Actually, if you haven't tried it since version 5 or 6, you'll find the rendering engine is as good as mozilla, and occasionally better. Its now a case of specific site- variation, rather than one being obviously better. The main re-write in this version has been the CSS engine - which is superb now, and renders anything I throw at it.
The main thing I like about Opera though is the mouse gestures - now customisable. (I know, Mozilla has them too, but I still find Opera a lot more responsive, and nicer to use.
Isn't that the point of remedies to illegal usage of monopolies? to help the people damaged by that monopoly?
When they do have implementation specific things, they try to put in a range. They teach ML and Java, and insist that the excercises for one are written in windows, and the other in unix, and so on. They also have comparative {OS,programming languages,etc} courses.
The aim is to produce people who aren't tied to one thing - since that will probably have changed while they are doing their undergrad, but can adapt.
The discrimination might not be an issue, but its still a huge privacy violation!
global tlds - require some sort of evidence of being a global company (there are country codes for a reason)
and finally a tld where there are no such restrictions, but it is strictly first come first serve - if I want to register microsoft.whatever, fine - and it can't be taken back for using their name, companies have .com and .co.cc and so on.
Oh, and IE 5.0 did complain about the certificate being invalid
The problem you have with this is not the signing itself, but with key trust relationships. Basically I can create a key with any name & email address - how do you know that it is actually me? There are several methods to solve this. Either you have to meet in person and give them the key there, or it has to be signed by a mutually trusted 3rd party. Meeting in person usually defies the point of public key encryption - you might as well just use symmetric if you can give them the key. The second option is a trusted 3rd party - but who do you trust that much, if you recall there was a slashdot story a while back (can't find the link) about Verisign signing dodgy Microsoft certificates.
Gnupg actually uses a web of trust system - if you have verified someone's key in person, and you have signed it to say you trust it, then it assumes you trust any key they sign. This introduces another weak point though. Even if you only sign keys after personally verifying fingerprints, do all the people who you've signed?
There are several program out there you can use, but if you are interfacing with customers using this, either you verifying their signatures, or they checking yours - there are going to be 2 problems. Do they trust the tech in general, and are they prepared to trust it, and how do you trust each other's keys. There probably needs to be a high profile 3rd party that you can both trust to authenticate against - but are there any at the moment for use with general signing tools? (Something like the certificates used in https/SSL, probably)
Basically, there is one system that holds all the data about which flight is assigned to which transponder ID number, and what air routes etc. This is a big mainframe, currently running on 3 IBM S390s (I saw them when they were putting them in last time I was in there). The actual code, however, is the same as they were using on the original S320, and has survived 2 replacement systems, by the cunning use of emulators. Incidentally the code is so old, it doesn't store dates (and hence survived the millenium bug) - it just rolls over at midnight - all the numbers get reset / reassigned.
This code crashes fairly regularly, but until now wasn't a serious problem - because they could just restart it. Unfortunately, the new centre in Swanwick produced some new issues. The new centre takes a data feed from the mainframe at the old site. The designers of the 2 systems didn't consider the fact that the two might become out of sync, and hence when the mainframe gets reset, they go to manual in the interim, and the 2 systems are out of sync when they come back up. At one point they had to wait until midnight, reset both systems (because everything rolls over at midnight) and go from there. They appear to be able to bring the systems back in sync, and back to normal speed, within a few hours now.
I suspect the software upgrade was to fix this problem, but didn't go according to plan.
Of course the new centre itself is another story. As pointed out there were some bad design decisions (The American's - who were going to make it cheaper by doing all the devopment - they stopped, lets carry on!!) Plus the fact that Lockheed Martin were going to walk out on the contract, because the penalty clause was going to be less than the cost of finishing the project....