If you think that adding one line to the boot.ini text file is 'fucking with my bootloader', I think you should stay well away from anything like installing an operating system anyway, and consider removing all the scissors from your house just in case you have an accident with them too.
Security has a lot to do with usability - one example: having cleaned up quite a few Windows PCs that had become almost unusable due to drive-by malware downloads installing unwanted toolbars, with pop-up windows everywhere and incredibly slow performance, I'd say that a secure system will not suffer this sort of degradation of usability that's common to Windows PCs not run by techies.
On the 'risk my Vista install' on your desktop - you can now try out Ubuntu with zero risk by installing it to a large file that's entirely within the Windows filesystem. The installer is just a few options on a single page. The only change it makes is to your c:\boot.ini file, and installing a boot program underneath the standard Windows Vista boot loader (NTLDR) - when you reboot, you can choose Ubuntu and it boots from the large file that contains the Ubuntu system. There is no repartitioning involved so this is no more risky than installing a Windows app, and probably less so as it does nothing to your registry etc.
KDE4 really needs a HUGE amount of usability testing before it is finalised - I tried the Kubuntu KDE4 live CD, and the whole K menu is really very unusable, in the way that it's stateful, remembering the last submenu you were in every time you click the K button again...
GNOME is the way it is today largely because of usability testing, I believe - while many power users and developers whinge about this, it is becoming much closer to Macs in overall usability.
So the issue is not "stupid developers", it's a matter of taking the time to do the testing - and it helps if you have some expertise at running the tests. Then it's the time to actually make the changes. Many developers aren't that interested in doing the testing, which is why there have been separate usability initiatives that can feed changes into projects.
Some of the issues logged here are not that easy to solve - e.g. making Firefox pop up an Ubuntu-specific Flash installation prompt, rather than executing the YouTube JavaScript logic that pushes people towards an Adobe plugin site that actually does have a Linux plugin for Flash, but one that's much harder to install than an Ubuntu-packaged Flash plugin.
Also, the one about finding MP3s on the Windows partition is not that easy - you could simply copy the files across with the Ubuntu migration assistant, but what if they're in a non-standard place? Indexing the Windows filesystem to quickly find these might help, but building the index could take some time. However, it would probably be enough if there was some feature in Ubuntu that scanned for existing partitions and said (based on partition type and a few key directories/files) that 'this looks like a Windows partition, it's available on the desktop through this icon', and ideally did a special symbolic link for the My Documents or similar (though that's tough as it's per-user under Windows - which user should this use).
Actually the eee community has discussed this many times, and the flash world has examined this in depth - it seems that with realistic usage patterns and reasonably good wear levelling there's much less risk of flash wearing out (see http://forum.eeeuser.com/viewtopic.php?id=7077 and in particular the link in first post to http://storagesearch.com/ ).
However, a system that minimises flash writes will also perform much better, since flash is rather slow at present for writing compared to hard disk or RAM.
I've been reading up a lot on flash drive technology recently, and it's seems that Xandros on the eee has been tuned somewhat to run well on flash (unionfs, run mostly in RAM, etc) to ensure that not to many writes are made to the flash drive. Generally most flash today is NAND based and has 100K write/erase cycles - some embedded-quality industrial flash drives have better ECC, wear levelling and bad block management to go somewhat higher (but you then pay more for the CF or SD card) - so it's important to do this to extend flash drive's lifetime. However the trend is for low-end flash to use MLCs (>1 bit per cell, vs SLCs which have 1 bit per cell) - drives using MLCs typically have even lower flash lifetime (10K write cycles), and the flash drive manufacturers are usually vague on this, particularly the cheaper ones.
The write cycles are across each individual erase block (something like 32 to 128 Kbyte), not per sector/page. Bad block management is critical to 'wear levelling' - as one erase block gets worn out (flagged by ECC) the data is moved across to a new erase block. As long as there are enough good erase blocks and you aren't doing a lot of writes to every part of the drive, there should be enough good blocks around to substitute for bad blocks. There's also work to ensure that if power is lost while multiple pages are written to an erase block, the drive can detect which were written OK - it then reads these and writes them to a new erase block, marking the old erase block as bad. The flash drive has a software Flash Translation Layer (FTL) that hides all this complexity, and the better vendors put more effort into good FTLs.
So... Some care is needed to install another Linux distro, or standard XP, onto the eee - not to get it installed, but to avoid wearing out the eee's flash drives too quickly. There are various flash-optimised Linux distros including Damn Small Linux (DSL, http://damnsmalllinux.org/ Puppy, SLAX, Debian Live (http://debian-live.alioth.debian.org/), etc, which manage to write infrequently to flash by running from a RAM disk (with no swap on flash, or at least reduced 'swappiness' parameter) and using unionfs or aufs to map a RAM drive 'over' the flash drive, allowing writes to be delayed until much later, and thereby minimising number of flash writes. DSL writes only when you shut down, or on demand, and Puppy writes every 30 minutes or so. Generally, Live CD distros are quite easily adapted to run well with flash, whereas hard disk distros do not run well on flash.
Ubuntu for eee looks very nice if you like Ubuntu, but doesn't do any flash optimisation that I could see from its wiki (apart from recommending use of noatime in fstab which is quite basic) - perhaps someone has done this as an add-on though. XP embedded apparently has some tweaks to do the same thing as Linux, but you need to be quite a techie to find and apply the flash optimisations, compared to simply installing Damn Small Linux which is already flash optimised.
There seems to be a lot of confusion on this - a good summary of this from eee perspective is http://bbs.archlinux.org/viewtopic.php?pid=297632. However, some people actually advocate removing unionfs from the eee Xandros setup in order to gain some flexibility, without even mentioning the issue of increased flash wear - see http://wiki.eeeuser.com/howto:removeunionfs which also suggests use of ext3 which will further increase flash writes (default is to write to log every 5 seconds typically). This is a really bad idea... I would really suggest reading up on this before changing the default setup, which uses unionfs in a similar way to DSL and Puppy Linux to minimise flash writes.
Does anyone know a major distro that runs on the eee and is already flash optimised to minimise writes?
Your experience is unfortunate, but it's really not that representative - the vast majority of Ubuntu installs go cleanly, as with other distros, but there are always cases where distro 1 fails to install easily, while distro 2 is fine, and others where this is reversed. The variability is driven by variation in hardware, which similar to Windows XP/Vista installs in that third party drives for those OSs can be hard to locate, have problems, etc.
Once you have Linux installed, which is by far the hardest part, it's very easy to keep it working, with no need for add-on products such as anti-spyware (I've tried several products for this including one that caused many false positives, another that didn't find some annoying spyware, another that worked well but required many add-on tools to erase one nasty spyware, etc), anti-virus, Secunia PSI to scan third party apps, firewall, defragmenter, commercial backup tool, Microsoft update (which sometimes takes 100% CPU, so I had to download a special patch), and so on, and on, and on. Just maintaining my Windows laptop for work, and my Windows home PC, takes far more effort than managing my three Linux boxes.
Try installing Ubuntu on five different computers, then you might have something more useful to report. I've installed it on two computers without problems, and the Hardy install (8.04 beta) was much better at recognising the video hardware that Ubuntu 5.10 and 7.04 stumbled on, so it's going in the right direction. Someone at work installed it on a PC simply as a way of getting a reliable X server (compared to Windows) for access to a Solaris server, and they're very happy with it. Another colleague has installed it on one of his home PCs and likes it a lot, including running some Windows apps on Linux via Crossover.
Try Damn Small Linux - it lets you use apt-get, and is based on Debian plus Knoppix, although it's not exactly a standard Debian setup due to the need to work well in a Flash or live CD environment, so there's heavy use of union filesystems and RAM disks. See my other comment at http://linux.slashdot.org/comments.pl?sid=512126&cid=22976596 for details.
I've been testing Puppy and Damn Small Linux (DSL, http://damnsmalllinux.org/) recently on a Thinkpad 560 with 96 MB RAM. Both are well optimised for booting from flash drives. Even if you have a PC that won't boot from Flash normally, you can still use Flash either with a boot floppy (WakePup for Puppy, DSL Boot USB for DSL), or with a Compact Flash card if it's a laptop - this appears as an IDE disk even to the BIOS, typically. Compact Flash should be faster than a USB drive and it's very easy to get a CF to PCMCIA connector (try hdparm -tT/dev/sdX1 to test read speeds).
Both work very fast - Puppy is better for Windows refugees who haven't yet learnt much about Linux, and is a bit more fragmented (it has 3 software installation mechanisms, some by the community and one by the author), and seemed less stable (only the package installation process though). The forums are very active but hosted on a slow server, and the Wiki is not too complete. Generally I found it hard to get the best information as there are many related websites that each have part of the story, but Puppy is very impressive and is particularly good on flash drives. Puppy has many different variants as well as the official ones, which can make it hard to work out which one to use, but that shows it has a very active community. However there is only one core developer.
DSL feels slicker in some ways but has less eye candy - it uses Knoppix hardware detection, which is of course excellent, although in my case I had to play with settings to get X11 to work. You can even use apt-get (with some limitations) to install Debian Woody. It works very well on Flash drives (USB or CF) - as with Puppy and many other live CDs, you can do a 'frugal' install in which a single file hosts a loop filesystem, minimising the writes to the Flash and thus improving lifetime of your Flash drive (a big issue as a given flash drive 'erase block' lasts only 10,000 to 100,000 writes). Like Puppy, it has a good mechanism for automatically saving your configuration and other RAM disk based state into the flash drive, though Puppy is slightly more automatic. Also, DSL includes vim by default, and is more focused on good command line tools as well as GUI tools, e.g. you can easily upgrade from BusyBox tools to GNU tools. Puppy tries to do everything through the GUI, so the vi sucks (actually e3vi, a very incomplete emulation).
Basically, if you are a Linux newbie, use Puppy, but if you know a little more about Linux already, use DSL. Since each distro will take quite a lot of configuration and learning about how it does things (quite different to Fedora or Ubuntu), it's best to choose the one you feel most comfortable with for longer term.
WiMAX is not magical pixie dust that you can sprinkle around to create a fabulous wireless service. It requires careful RF engineering - as other have pointed out, at its most common frequencies it doesn't have good non-line-of-sight support, but then neither do other protocols in those frequencies (and LTE, the 4G development from the 3GPP people who defined UMTS/3G and GSM, is based on very similar technology). Indoor performance is gated by quality of antenna - I understand Buzz are using antennas on the modem rather than mounting them on the house, which is a guarantee of problems with indoor coverage, not surprisingly.
There are many operators using WiMAX for VoIP and getting good RF performance (i.e. fewer retries and less jitter). It sounds like Buzz needs some serious work on their RF and network engineering, which is hardly the fault of WiMAX...
Incidentally, top tier vendors are all supporting WiMAX to cover the risk that WiMAX becomes a big market - they can't afford to be left out. Nokia, Samsung, Motorola, Nortel and others are doing WiMAX user devices and network equipment - Ericsson is not but that's because they don't want to cannibalise their huge share of the UMTS/HSPA 3G/3.5G market, and future LTE market. LTE will also be huge, and may still win out over WiMAX, but it won't be due to fundamental issues with WiMAX technology - most likely WiMAX will be strongest outside the traditional mobile/cellular operators, with LTE dominating within such operators.
In my view, this article is really about a clueless CEO trying to throw mud at WiMAX to distract investors from the fact that his company has messed up their network design and deployment.
Enabling IPv6 in Firefox only 'murders your browsing speed' if (most likely) you are served a AAAA DNS record for a site AND you have an IPv6 interface enabled on your machine AND it's connected to a broken IPv6 network - the OS (whether Windows or Linux) will try to use IPv6, then fail after a timeout, and maybe revert to IPv4 if you are lucky. So it's not Firefox or even the OS, it's having an IPv6 interface that is up while the IPv6 network you are on is down. This is rather a long way from being a Firefox problem.
The only other reason I can think of for IPv6 being slower is if you are tunnelling IPv6 over IPv4 to a distant tunnel broker node. This too is nothing to do with Firefox or even your machine.
I tried Firefox 3 (previous beta) on a 7 year old PC recently - Pentium III 700 MHz, 512 MB - and it ran *much* faster than Firefox 2 on many operations, particularly hitting the Back button. It also used less memory over time, as Mozilla have fixed many, many memory leaks.
I used it as my day to day browser - the only thing stopping me upgrading completely were a few extensions that aren't yet migrated to Firefox 3 (primarily Sage RSS reader - no longer maintained so I'll use Wizz instead).
That's the wrong link - you are on the Kubuntu site (where wiki.kubuntu.org maps to wiki.ubuntu.org it seems) but on the Ubuntu Hardy Alpha 4 page. Try https://wiki.kubuntu.org/HardyHeron/Alpha4/Kubuntu instead which talks about new KDE stuff in Alpha 4.
I've never found Ubuntu failing to boot at all over several systems and releases since 2005. I've also burnt CDs for several friends and they found it worked perfectly out of the box. Hardware support is variable, and I really wish Ubuntu would steal the great hardware *detection* from Knoppix, but generally Ubuntu is pretty good - some distros will do better on some machines, others worse, as you'd expect.
I do find Ubuntu is not good at handling machines with two video cards, with the motherboard card having no monitor attached - requires alternate installer and some tweaking. However that may be my fault for not disabling on-board video in the BIOS.
I'd imagine that the Desktop Team in Canonical asked the KDE developers about 3.5.x support when deciding this. Since Ubuntu is primarily GNOME-focused, I'd guess that the volume of KDE-based support contracts is not as high, which will tend to determine whether they can fund KDE-only developers to focus on 3.5.x support, or at least do it part time along with KDE 4.0 work. Personally, I use Kubuntu, but I don't need a support contract so it mostly affects LTS for commercial users.
In any case, 8.04 will include 3.5.x, but the Kubuntu 8.04 release as a whole won't be LTS since it will change to KDE 4.0 or 4.1 (I'd guess in 8.10).
From what I've seen of KDE 4.0 using the Kubuntu Live CD, it has some great ideas and features but is nowhere ready for prime time in various areas, so I'd prefer for 8.10 to be the LTS with KDE 4.x.
Actually it's pretty complex for a GUI tool to figure out actual disk impact of its operations - it needs to consider
- which volumes are mounted under which mount points (quite easy, but then calculate that for all files transferred, not just root of each transfer)
- which volumes are really LVM2 logical volumes, and how those are mapped to volume groups and then physical volumes (LVM is default in Fedora and maybe also Ubuntu, I always install it explicitly)
- is RAID in use, at level of software RAID (Linux only), fake RAID (hardware assist + Linux) or hardware RAID, and if so are you using RAID 0, 1, 5, 6, 10, etc, and which blocks might be on which disks given different RAID striping models and which disks are in which group. (And once we get ZFS-FUSE to be stable, that has its own version of RAID...)
- is this disk actually mapped at block level onto a network block device (e.g. iSCSI, ATA over Ethernet, etc - or even ElasticDrive which is really an Amazon S3 based storage service)
Given LVM and RAID in particular, I can't see how any GUI tool can figure this out easily - it would really need to poke the kernel to get some hints if you want to do this, or perhaps hint to the kernel 'these 5 operations are a group, please schedule them sequentially if that will improve performance' - still very complex however you do it.
One interesting feature here, to set priorities for I/O between processes although not threads, is 'ionice' - it's available in Ubuntu from repositories and let's you set the priority for I/O of a process - great for disk-bound transfers, e.g. set your K3B disk burning to high priority, or backup process to low. Useful blog posting at http://friedcpu.wordpress.com/2007/07/17/why-arent-you-using-ionice-yet/
Incidentally Vista added I/O priority to Windows, but XP doesn't have it, which is why I need to suspend some processes altogether when they are churning the disk on my Windows laptop.
I would like to see the various GUI system monitor tools make it easy to see how much I/O is going on, like Windows' Process Explorer (from the great Sysinternals team, far better than Task Manager) and to view and change the I/O priority interactively.
Having said all that - I think Nautilus is going to include a Pause transfer feature in next iteration, so you could simply use that to stop something that may be slowing things down.
I've not had any problems upgrading from Feisty to Gutsy on laptop, and generally the Ubuntu forums are amazing - I've not had to post questions yet, simply because the volume of traffic and useful answers means there is almost always a good answer findable using Google.
The one thing that would help is making the Ubuntu wiki more search engine friendly, and more actively integrating the wiki and forums - why can't a HOWTO post seamlessly turn into a Wiki page while also remaining the top post of the HOWTO thread? There's a lot of good content on the Wiki but it rarely turns up in Google, perhaps because they use SSL for all pages. If only they used TWiki as a base, which even includes AJAX-based Google searching within the TWiki site as well as WYSIWYG editing, revision control, 400 plugins...
This is complete crap - not only is Ubuntu 100% open source, with standard source packages, it is the root of many other distributions and almost encourages people to fork Ubuntu - e.g. Mythbuntu, Ubuntu Studio, Linux Mint, Fluxbuntu, and more. They do release their own contributions and actively merge them upstream into Debian, so they are also helping every other Debian-based distro (e.g. Sidux, Knoppix, SimplyMEPIS and many others).
I used the one from Allendale in the UK: http://www.wifi-antennas.co.uk/index.php?target=products&product_id=15 - not sure of part no. but you could ask them. They have some great FAQs and guides, but fundamentally the best thing to do is buy a 9 dBi antenna and see if it works for you, as it costs very little. Be sure to check your router's antenna connection and note that the very latest (crap) WRT54G version doesn't have a removable antenna...
Which reminds me, Netgear routers seem to redirect 192.168.0.1 to 'routerlogin.com' (owned by Netgear, but actually maps to your router normally). A somewhat dodgy design decision really as it serves to obfuscate what's happening when newbies log on to their router, which can't help them to learn more about security.
You are confusing static addressing (which doesn't help) with choosing a less obvious IP range (which does). It's fine to use DHCP, just change the address you use to something non-default - anything in 192.168.0.0/24 or 172.x or 10.x ranges (check RFC1918) is good. That way the malware on a PC will have to scan first to know what to attack, which raises the bar slightly, although at Ethernet speeds it wouldn't take very long to scan the whole 192.168.x address range.
BTW the 'third and fourth hex digits' doesn't make sense - I guess you mean the third and fourth octets e.g. 205 and 89 in your example IP address, or the lower 16 bits of the IPv4 address.
Incidentally, IPv6 will be more secure in this respect since it uses the MAC address (or a random number) for the lowest 64 bits of the 128 bit address.
Having said all that, I just did a test of "ping 192.168.x.255" for my home LAN, and I got a nice response from only one device - the home router (DD-WRT on WRT54G)... My Ubuntu box and a Windows box didn't respond, interestingly - apparently Windows never responds to broadcast pings. Pinging 192.168.255.255 proved only that my home router and ISP don't do bogon filtering as it reached the ISP's router. Oh well...
Since broadcast pings work fine, the only question becomes how to write malware that can do a ping for this simple network discovery - the answer is a signed ActiveX control or Java applet, which is how most spyware gets installed, so that isn't too hard.
There's no really secure way to do this - DD-WRT v24 is supposed to enable multiple simultaneous SSIDs, some locked (WPA) and some not, but unfortunately it's still quite unstable. Since I discovered that WDS on DD-WRT v23 is really quite unstable, I've been considering moving to another Linux-based firmware (maybe Tomato or OpenWRT with add-on GUI), but currently I don't need WDS since I bought a 9 dBi omnidirectional antenna for $25. I'd recommend such antennas to anyone before wasting time with 802.11n, WDS, 802.11g+MIMO, etc - I now have coverage of a large house with 2 foot thick stone walls from a single WRT54G at recommended max power of 84 mW. I now use the WDS router as a spare.
You really need to be more specific than 'serious pain in the butt'... are you talking about bugs, performance issues, or what?
I don't think anyone is intending NTFS-3G to be used on a server, since it's really for dual-boot which is mostly for desktops and laptops - just use ext3 and Samba. And I think there are enough desktop users of NTFS-3G that I can't see it causing problems for work usage.
Slashdot Mirror
If you think that adding one line to the boot.ini text file is 'fucking with my bootloader', I think you should stay well away from anything like installing an operating system anyway, and consider removing all the scissors from your house just in case you have an accident with them too.
Security has a lot to do with usability - one example: having cleaned up quite a few Windows PCs that had become almost unusable due to drive-by malware downloads installing unwanted toolbars, with pop-up windows everywhere and incredibly slow performance, I'd say that a secure system will not suffer this sort of degradation of usability that's common to Windows PCs not run by techies.
On the 'risk my Vista install' on your desktop - you can now try out Ubuntu with zero risk by installing it to a large file that's entirely within the Windows filesystem. The installer is just a few options on a single page. The only change it makes is to your c:\boot.ini file, and installing a boot program underneath the standard Windows Vista boot loader (NTLDR) - when you reboot, you can choose Ubuntu and it boots from the large file that contains the Ubuntu system. There is no repartitioning involved so this is no more risky than installing a Windows app, and probably less so as it does nothing to your registry etc.
The tool (Wubi) that does this is now part of Ubuntu 8.04, see https://wiki.ubuntu.com/WubiGuide for more information.
KDE4 really needs a HUGE amount of usability testing before it is finalised - I tried the Kubuntu KDE4 live CD, and the whole K menu is really very unusable, in the way that it's stateful, remembering the last submenu you were in every time you click the K button again...
Actually usability testing has been going on with Linux for many years - since at least 2001 for GNOME when Sun started doing this ( http://developer.gnome.org/projects/gup/usertesting.html ). Here's a good article that talks about usability testing for Linux, also from 2001, and mentioning KDE user testing: http://lwn.net/2001/0614/desktop.php3
GNOME is the way it is today largely because of usability testing, I believe - while many power users and developers whinge about this, it is becoming much closer to Macs in overall usability.
So the issue is not "stupid developers", it's a matter of taking the time to do the testing - and it helps if you have some expertise at running the tests. Then it's the time to actually make the changes. Many developers aren't that interested in doing the testing, which is why there have been separate usability initiatives that can feed changes into projects.
Some of the issues logged here are not that easy to solve - e.g. making Firefox pop up an Ubuntu-specific Flash installation prompt, rather than executing the YouTube JavaScript logic that pushes people towards an Adobe plugin site that actually does have a Linux plugin for Flash, but one that's much harder to install than an Ubuntu-packaged Flash plugin.
Also, the one about finding MP3s on the Windows partition is not that easy - you could simply copy the files across with the Ubuntu migration assistant, but what if they're in a non-standard place? Indexing the Windows filesystem to quickly find these might help, but building the index could take some time. However, it would probably be enough if there was some feature in Ubuntu that scanned for existing partitions and said (based on partition type and a few key directories/files) that 'this looks like a Windows partition, it's available on the desktop through this icon', and ideally did a special symbolic link for the My Documents or similar (though that's tough as it's per-user under Windows - which user should this use).
Actually the eee community has discussed this many times, and the flash world has examined this in depth - it seems that with realistic usage patterns and reasonably good wear levelling there's much less risk of flash wearing out (see http://forum.eeeuser.com/viewtopic.php?id=7077 and in particular the link in first post to http://storagesearch.com/ ).
However, a system that minimises flash writes will also perform much better, since flash is rather slow at present for writing compared to hard disk or RAM.
I've been reading up a lot on flash drive technology recently, and it's seems that Xandros on the eee has been tuned somewhat to run well on flash (unionfs, run mostly in RAM, etc) to ensure that not to many writes are made to the flash drive. Generally most flash today is NAND based and has 100K write/erase cycles - some embedded-quality industrial flash drives have better ECC, wear levelling and bad block management to go somewhat higher (but you then pay more for the CF or SD card) - so it's important to do this to extend flash drive's lifetime. However the trend is for low-end flash to use MLCs (>1 bit per cell, vs SLCs which have 1 bit per cell) - drives using MLCs typically have even lower flash lifetime (10K write cycles), and the flash drive manufacturers are usually vague on this, particularly the cheaper ones.
The write cycles are across each individual erase block (something like 32 to 128 Kbyte), not per sector/page. Bad block management is critical to 'wear levelling' - as one erase block gets worn out (flagged by ECC) the data is moved across to a new erase block. As long as there are enough good erase blocks and you aren't doing a lot of writes to every part of the drive, there should be enough good blocks around to substitute for bad blocks. There's also work to ensure that if power is lost while multiple pages are written to an erase block, the drive can detect which were written OK - it then reads these and writes them to a new erase block, marking the old erase block as bad. The flash drive has a software Flash Translation Layer (FTL) that hides all this complexity, and the better vendors put more effort into good FTLs.
So... Some care is needed to install another Linux distro, or standard XP, onto the eee - not to get it installed, but to avoid wearing out the eee's flash drives too quickly. There are various flash-optimised Linux distros including Damn Small Linux (DSL, http://damnsmalllinux.org/ Puppy, SLAX, Debian Live (http://debian-live.alioth.debian.org/), etc, which manage to write infrequently to flash by running from a RAM disk (with no swap on flash, or at least reduced 'swappiness' parameter) and using unionfs or aufs to map a RAM drive 'over' the flash drive, allowing writes to be delayed until much later, and thereby minimising number of flash writes. DSL writes only when you shut down, or on demand, and Puppy writes every 30 minutes or so. Generally, Live CD distros are quite easily adapted to run well with flash, whereas hard disk distros do not run well on flash.
Ubuntu for eee looks very nice if you like Ubuntu, but doesn't do any flash optimisation that I could see from its wiki (apart from recommending use of noatime in fstab which is quite basic) - perhaps someone has done this as an add-on though. XP embedded apparently has some tweaks to do the same thing as Linux, but you need to be quite a techie to find and apply the flash optimisations, compared to simply installing Damn Small Linux which is already flash optimised.
There seems to be a lot of confusion on this - a good summary of this from eee perspective is http://bbs.archlinux.org/viewtopic.php?pid=297632. However, some people actually advocate removing unionfs from the eee Xandros setup in order to gain some flexibility, without even mentioning the issue of increased flash wear - see http://wiki.eeeuser.com/howto:removeunionfs which also suggests use of ext3 which will further increase flash writes (default is to write to log every 5 seconds typically). This is a really bad idea... I would really suggest reading up on this before changing the default setup, which uses unionfs in a similar way to DSL and Puppy Linux to minimise flash writes.
Does anyone know a major distro that runs on the eee and is already flash optimised to minimise writes?
Your experience is unfortunate, but it's really not that representative - the vast majority of Ubuntu installs go cleanly, as with other distros, but there are always cases where distro 1 fails to install easily, while distro 2 is fine, and others where this is reversed. The variability is driven by variation in hardware, which similar to Windows XP/Vista installs in that third party drives for those OSs can be hard to locate, have problems, etc.
Once you have Linux installed, which is by far the hardest part, it's very easy to keep it working, with no need for add-on products such as anti-spyware (I've tried several products for this including one that caused many false positives, another that didn't find some annoying spyware, another that worked well but required many add-on tools to erase one nasty spyware, etc), anti-virus, Secunia PSI to scan third party apps, firewall, defragmenter, commercial backup tool, Microsoft update (which sometimes takes 100% CPU, so I had to download a special patch), and so on, and on, and on. Just maintaining my Windows laptop for work, and my Windows home PC, takes far more effort than managing my three Linux boxes.
Try installing Ubuntu on five different computers, then you might have something more useful to report. I've installed it on two computers without problems, and the Hardy install (8.04 beta) was much better at recognising the video hardware that Ubuntu 5.10 and 7.04 stumbled on, so it's going in the right direction. Someone at work installed it on a PC simply as a way of getting a reliable X server (compared to Windows) for access to a Solaris server, and they're very happy with it. Another colleague has installed it on one of his home PCs and likes it a lot, including running some Windows apps on Linux via Crossover.
Try Damn Small Linux - it lets you use apt-get, and is based on Debian plus Knoppix, although it's not exactly a standard Debian setup due to the need to work well in a Flash or live CD environment, so there's heavy use of union filesystems and RAM disks. See my other comment at http://linux.slashdot.org/comments.pl?sid=512126&cid=22976596 for details.
I've been testing Puppy and Damn Small Linux (DSL, http://damnsmalllinux.org/) recently on a Thinkpad 560 with 96 MB RAM. Both are well optimised for booting from flash drives. Even if you have a PC that won't boot from Flash normally, you can still use Flash either with a boot floppy (WakePup for Puppy, DSL Boot USB for DSL), or with a Compact Flash card if it's a laptop - this appears as an IDE disk even to the BIOS, typically. Compact Flash should be faster than a USB drive and it's very easy to get a CF to PCMCIA connector (try hdparm -tT /dev/sdX1 to test read speeds).
Both work very fast - Puppy is better for Windows refugees who haven't yet learnt much about Linux, and is a bit more fragmented (it has 3 software installation mechanisms, some by the community and one by the author), and seemed less stable (only the package installation process though). The forums are very active but hosted on a slow server, and the Wiki is not too complete. Generally I found it hard to get the best information as there are many related websites that each have part of the story, but Puppy is very impressive and is particularly good on flash drives. Puppy has many different variants as well as the official ones, which can make it hard to work out which one to use, but that shows it has a very active community. However there is only one core developer.
DSL feels slicker in some ways but has less eye candy - it uses Knoppix hardware detection, which is of course excellent, although in my case I had to play with settings to get X11 to work. You can even use apt-get (with some limitations) to install Debian Woody. It works very well on Flash drives (USB or CF) - as with Puppy and many other live CDs, you can do a 'frugal' install in which a single file hosts a loop filesystem, minimising the writes to the Flash and thus improving lifetime of your Flash drive (a big issue as a given flash drive 'erase block' lasts only 10,000 to 100,000 writes). Like Puppy, it has a good mechanism for automatically saving your configuration and other RAM disk based state into the flash drive, though Puppy is slightly more automatic. Also, DSL includes vim by default, and is more focused on good command line tools as well as GUI tools, e.g. you can easily upgrade from BusyBox tools to GNU tools. Puppy tries to do everything through the GUI, so the vi sucks (actually e3vi, a very incomplete emulation).
Basically, if you are a Linux newbie, use Puppy, but if you know a little more about Linux already, use DSL. Since each distro will take quite a lot of configuration and learning about how it does things (quite different to Fedora or Ubuntu), it's best to choose the one you feel most comfortable with for longer term.
WiMAX is not magical pixie dust that you can sprinkle around to create a fabulous wireless service. It requires careful RF engineering - as other have pointed out, at its most common frequencies it doesn't have good non-line-of-sight support, but then neither do other protocols in those frequencies (and LTE, the 4G development from the 3GPP people who defined UMTS/3G and GSM, is based on very similar technology). Indoor performance is gated by quality of antenna - I understand Buzz are using antennas on the modem rather than mounting them on the house, which is a guarantee of problems with indoor coverage, not surprisingly.
There are many operators using WiMAX for VoIP and getting good RF performance (i.e. fewer retries and less jitter). It sounds like Buzz needs some serious work on their RF and network engineering, which is hardly the fault of WiMAX...
Incidentally, top tier vendors are all supporting WiMAX to cover the risk that WiMAX becomes a big market - they can't afford to be left out. Nokia, Samsung, Motorola, Nortel and others are doing WiMAX user devices and network equipment - Ericsson is not but that's because they don't want to cannibalise their huge share of the UMTS/HSPA 3G/3.5G market, and future LTE market. LTE will also be huge, and may still win out over WiMAX, but it won't be due to fundamental issues with WiMAX technology - most likely WiMAX will be strongest outside the traditional mobile/cellular operators, with LTE dominating within such operators.
In my view, this article is really about a clueless CEO trying to throw mud at WiMAX to distract investors from the fact that his company has messed up their network design and deployment.
Enabling IPv6 in Firefox only 'murders your browsing speed' if (most likely) you are served a AAAA DNS record for a site AND you have an IPv6 interface enabled on your machine AND it's connected to a broken IPv6 network - the OS (whether Windows or Linux) will try to use IPv6, then fail after a timeout, and maybe revert to IPv4 if you are lucky. So it's not Firefox or even the OS, it's having an IPv6 interface that is up while the IPv6 network you are on is down. This is rather a long way from being a Firefox problem.
The only other reason I can think of for IPv6 being slower is if you are tunnelling IPv6 over IPv4 to a distant tunnel broker node. This too is nothing to do with Firefox or even your machine.
I tried Firefox 3 (previous beta) on a 7 year old PC recently - Pentium III 700 MHz, 512 MB - and it ran *much* faster than Firefox 2 on many operations, particularly hitting the Back button. It also used less memory over time, as Mozilla have fixed many, many memory leaks.
I used it as my day to day browser - the only thing stopping me upgrading completely were a few extensions that aren't yet migrated to Firefox 3 (primarily Sage RSS reader - no longer maintained so I'll use Wizz instead).
That's the wrong link - you are on the Kubuntu site (where wiki.kubuntu.org maps to wiki.ubuntu.org it seems) but on the Ubuntu Hardy Alpha 4 page. Try https://wiki.kubuntu.org/HardyHeron/Alpha4/Kubuntu instead which talks about new KDE stuff in Alpha 4.
I've never found Ubuntu failing to boot at all over several systems and releases since 2005. I've also burnt CDs for several friends and they found it worked perfectly out of the box. Hardware support is variable, and I really wish Ubuntu would steal the great hardware *detection* from Knoppix, but generally Ubuntu is pretty good - some distros will do better on some machines, others worse, as you'd expect.
I do find Ubuntu is not good at handling machines with two video cards, with the motherboard card having no monitor attached - requires alternate installer and some tweaking. However that may be my fault for not disabling on-board video in the BIOS.
I'd imagine that the Desktop Team in Canonical asked the KDE developers about 3.5.x support when deciding this. Since Ubuntu is primarily GNOME-focused, I'd guess that the volume of KDE-based support contracts is not as high, which will tend to determine whether they can fund KDE-only developers to focus on 3.5.x support, or at least do it part time along with KDE 4.0 work. Personally, I use Kubuntu, but I don't need a support contract so it mostly affects LTS for commercial users.
In any case, 8.04 will include 3.5.x, but the Kubuntu 8.04 release as a whole won't be LTS since it will change to KDE 4.0 or 4.1 (I'd guess in 8.10).
From what I've seen of KDE 4.0 using the Kubuntu Live CD, it has some great ideas and features but is nowhere ready for prime time in various areas, so I'd prefer for 8.10 to be the LTS with KDE 4.x.
Actually it's pretty complex for a GUI tool to figure out actual disk impact of its operations - it needs to consider
- which volumes are mounted under which mount points (quite easy, but then calculate that for all files transferred, not just root of each transfer)
- which volumes are really LVM2 logical volumes, and how those are mapped to volume groups and then physical volumes (LVM is default in Fedora and maybe also Ubuntu, I always install it explicitly)
- is RAID in use, at level of software RAID (Linux only), fake RAID (hardware assist + Linux) or hardware RAID, and if so are you using RAID 0, 1, 5, 6, 10, etc, and which blocks might be on which disks given different RAID striping models and which disks are in which group. (And once we get ZFS-FUSE to be stable, that has its own version of RAID...)
- is this disk actually mapped at block level onto a network block device (e.g. iSCSI, ATA over Ethernet, etc - or even ElasticDrive which is really an Amazon S3 based storage service)
Given LVM and RAID in particular, I can't see how any GUI tool can figure this out easily - it would really need to poke the kernel to get some hints if you want to do this, or perhaps hint to the kernel 'these 5 operations are a group, please schedule them sequentially if that will improve performance' - still very complex however you do it.
One interesting feature here, to set priorities for I/O between processes although not threads, is 'ionice' - it's available in Ubuntu from repositories and let's you set the priority for I/O of a process - great for disk-bound transfers, e.g. set your K3B disk burning to high priority, or backup process to low. Useful blog posting at http://friedcpu.wordpress.com/2007/07/17/why-arent-you-using-ionice-yet/
Incidentally Vista added I/O priority to Windows, but XP doesn't have it, which is why I need to suspend some processes altogether when they are churning the disk on my Windows laptop.
I would like to see the various GUI system monitor tools make it easy to see how much I/O is going on, like Windows' Process Explorer (from the great Sysinternals team, far better than Task Manager) and to view and change the I/O priority interactively.
Having said all that - I think Nautilus is going to include a Pause transfer feature in next iteration, so you could simply use that to stop something that may be slowing things down.
I've not had any problems upgrading from Feisty to Gutsy on laptop, and generally the Ubuntu forums are amazing - I've not had to post questions yet, simply because the volume of traffic and useful answers means there is almost always a good answer findable using Google.
The one thing that would help is making the Ubuntu wiki more search engine friendly, and more actively integrating the wiki and forums - why can't a HOWTO post seamlessly turn into a Wiki page while also remaining the top post of the HOWTO thread? There's a lot of good content on the Wiki but it rarely turns up in Google, perhaps because they use SSL for all pages. If only they used TWiki as a base, which even includes AJAX-based Google searching within the TWiki site as well as WYSIWYG editing, revision control, 400 plugins...
This is complete crap - not only is Ubuntu 100% open source, with standard source packages, it is the root of many other distributions and almost encourages people to fork Ubuntu - e.g. Mythbuntu, Ubuntu Studio, Linux Mint, Fluxbuntu, and more. They do release their own contributions and actively merge them upstream into Debian, so they are also helping every other Debian-based distro (e.g. Sidux, Knoppix, SimplyMEPIS and many others).
How did this get moderated insightful?
See http://it.slashdot.org/comments.pl?sid=426748&cid=22150290 for pointers on the antenna I used.
I used the one from Allendale in the UK: http://www.wifi-antennas.co.uk/index.php?target=products&product_id=15 - not sure of part no. but you could ask them. They have some great FAQs and guides, but fundamentally the best thing to do is buy a 9 dBi antenna and see if it works for you, as it costs very little. Be sure to check your router's antenna connection and note that the very latest (crap) WRT54G version doesn't have a removable antenna...
Which reminds me, Netgear routers seem to redirect 192.168.0.1 to 'routerlogin.com' (owned by Netgear, but actually maps to your router normally). A somewhat dodgy design decision really as it serves to obfuscate what's happening when newbies log on to their router, which can't help them to learn more about security.
You are confusing static addressing (which doesn't help) with choosing a less obvious IP range (which does). It's fine to use DHCP, just change the address you use to something non-default - anything in 192.168.0.0/24 or 172.x or 10.x ranges (check RFC1918) is good. That way the malware on a PC will have to scan first to know what to attack, which raises the bar slightly, although at Ethernet speeds it wouldn't take very long to scan the whole 192.168.x address range.
BTW the 'third and fourth hex digits' doesn't make sense - I guess you mean the third and fourth octets e.g. 205 and 89 in your example IP address, or the lower 16 bits of the IPv4 address.
Incidentally, IPv6 will be more secure in this respect since it uses the MAC address (or a random number) for the lowest 64 bits of the 128 bit address.
Having said all that, I just did a test of "ping 192.168.x.255" for my home LAN, and I got a nice response from only one device - the home router (DD-WRT on WRT54G)... My Ubuntu box and a Windows box didn't respond, interestingly - apparently Windows never responds to broadcast pings. Pinging 192.168.255.255 proved only that my home router and ISP don't do bogon filtering as it reached the ISP's router. Oh well...
Since broadcast pings work fine, the only question becomes how to write malware that can do a ping for this simple network discovery - the answer is a signed ActiveX control or Java applet, which is how most spyware gets installed, so that isn't too hard.
There's no really secure way to do this - DD-WRT v24 is supposed to enable multiple simultaneous SSIDs, some locked (WPA) and some not, but unfortunately it's still quite unstable. Since I discovered that WDS on DD-WRT v23 is really quite unstable, I've been considering moving to another Linux-based firmware (maybe Tomato or OpenWRT with add-on GUI), but currently I don't need WDS since I bought a 9 dBi omnidirectional antenna for $25. I'd recommend such antennas to anyone before wasting time with 802.11n, WDS, 802.11g+MIMO, etc - I now have coverage of a large house with 2 foot thick stone walls from a single WRT54G at recommended max power of 84 mW. I now use the WDS router as a spare.
You really need to be more specific than 'serious pain in the butt' ... are you talking about bugs, performance issues, or what?
I don't think anyone is intending NTFS-3G to be used on a server, since it's really for dual-boot which is mostly for desktops and laptops - just use ext3 and Samba. And I think there are enough desktop users of NTFS-3G that I can't see it causing problems for work usage.