Those who have done tests have found that kernel preemption reduces average latencies rather than worst case latencies (e.g. when your audio drops out or window stutters you are seeing worst case latencies) at the cost of hurting throughput a tiny bit (so things actually take longer because they are interrupted more). In general, the low latency patches had a bigger effect than preemption which is why many of the distros do not ship with a kernel with preemption on by default. However it is useful as a debugging aid and may well be a stepping stone to something else.
I suspect you will find more of the cheats that Window's uses making their way into Linux desktop distros because the user perception is improved. Currently I'm willing to trade a bit of RAM for a faster startup but it's how to do that so I do not notice that you are doing so is the real trick.
I don't know if it's possible for all the icons (e.g. those in the panel, open office etc.) but for window decorations and nautilus just change the GNOME theme from the GNOME control panel to something like Traditional.
Taking a look at the torrent file seems to suggest that the DVD version is just an ISO and nothing else. Be careful folks - without *signed* MD5SUMs you can't be sure of what you've downloaded.
Apparently there are some systems that yum simply can't handle because it has to update the system while it is "online" (e.g. LVM). So it looks like the answer is "it depends on your set up".
See Seth Videl's post about it. My advice is to wait and see what the pitfalls are since there *will* be gotchas.
...hmm. advogato's being a bit strange today so let me post a quote:
Wrote up some not-yet-finished notes on how to yum update from FC1 to FC2 with relative ease. I'll post a link here when I'm happy. It's not a hard process and for most people it'll work fairly ok. For some people, however, for example, people using LVM, there are certain things yum just can't do, and there is no nice way around it on a running system. This is where anaconda is the only way to do it. Since it is running outside of your installed system it can muck with things w/o worrying about making its environment completely unusable.
My understanding is that the *Universities of Kent and Lancaster* are no longer providing a mirroring service. There will be some sort of mirroring service provided to JANET users by a different third party. I think there was a tendering process and JANET decided to go with someone else this year.
However, I'm not sure that it's clear whether the new mirror will:
Carry all the mirrors the old mirror.ac.uk service did
Provide the variety of protocols the old mirror did (http/ftp/rsync)
Will be accessible to non JANET connections
The last point is the real sticking issue. Can anyone else clarify things? Either way mirror.ac.uk (as it is currently) will be sorely missed. It's provided an extremely useful service over the years and I'm sure it's saved Swansea Uni a lot of transatlantic traffic over the years:)
Read this article on the hoped move to Users with least priviledge in Windows Longhorn - it's just too risky to leave people running with maximum power all the time.
You're right preempt is turned off in Mandrake's stock kernels but it doesn't make a difference in this case since preempt only affects average case latencies not the worst case ones in a 2.6 kernel according to benchmarks.
The one big problem I've found with sophos is that the scanning engine itself will only work for 3 months before it has to be updated. This alone is incredibly frustrating.
You probably mean supermount as opposed to automount. You will be pleased to hear that supermount is no longer used by default and the Redhat / Fedora way of mounting is now used. A quick word of warning though - this has been quite a buggy release...
As to the autorun thing - were you using an NT/2K/XP machine? I'm fairly sure it does what is supposed to on 9x machines where you are more at liberty to poke memory but I'm fairly sure this process failed on NT series.
Just grab CD 1 and do a minimal install. Assuming you know the nams of the packages you want, you can (after configuring it) use yum to network install the rest once your system is up.
I'm fairly sure it is possible to do such minimal installs with Mandrake too (I've never tried) - just make sure you install only the minimum and be prepared for more hastle.
Have you tried contacting them? Email webmaster [at] mandrakeclub [dot] com and it's more like to be resolved. Seriously don't hang around on this (I know from experience). If you just complain on Slashdot your account may never be activated...
So what do you think happens when that proxy needs to make to make a link to a connection beyond the confines of ja.net because the latest page you are fetching from Slashdot has changed? It sends a connection out through the ja.net network and off via the transatlantic link. ja.net then record the amount of data sent across it and bill your University appropriately. Just because the immediate connection from the university is a ja.net link does not mean your uni won't be charged (per the Mbit) once it crosses out of ja.net.
I think you will find that UK Universities DO pay bandwidth costs for traffic that goes outside of ja.net (maybe they are charged for ja.net traffic too these days). Prices are especially steep if you generate transatlantic traffic. Caching of data has become very important in an effort to cut costs and it's rare that you are allowed to make www connections without going through some sort of Uni proxy.
Well the problem is that not everyone starts out using MS software. You have already pointed out Emacs (which might predate Windows but do you have any idea how frustrating it is to press ctrl-w to delete a word and have it close a window instead?) but the fact of the matter is something like Mozilla's newsreader may have inherited it's keybindings from Netscape's newsreader. Those defaults simply won't change because those were the ones that were there first. If I'd been using an app for years the last thing I want is for the keybindings to change to something else just because the developer wanted to be compatible with the competition so clearly that's not what is going to happen.
Perhaps you are aware that Word's keybindings are not indentical to those of Word Perfect's (despite the later having a massive market share when Word initially came out). The trick is to provide a migration mode initially to help people settle in. Perhaps those doing the complaining could implement a "MS Word User" option in OpenOffice that binds keyboard settings to those of Word. In the end users will give up the bindings and switch to the new one but it would help the inital process. I believe MS may have done something similar to help Netscape users switch from NS3 to IE4...
... with support on a Redhat system it appears wise to pay for Enterprise Linux that will be supported for at least 5 years (http://www.redhat.com/software/rhelorfedora/). If you want to avoid a regular upgrade cycle with Redhat then you have to be prepared to pay (which seems fair to me since backporting costs increasingly more as time goes by). Whether five years is enough is another question altogether though...
You are absolutely right about the old RPM db being prone to becoming corrupt in releases earlier than Fedora Core 1 and more than a few people have been burned by this. After Alan ran into such a corruption on one the Redhat 9 + ximain machines he was updating on SUCS he mentioned that such things should be a thing of the past from Fedora Core 1 onwards as RPM had been made transactional...
By the by, I have seen redhat be updated from 7.2 to Fedora Core 1 (stopping by 8.0, and 9) using up2date over a network connection (I don't know if this is approved either but it's all much of a muchness now unless you are using WS or ES). I think I have also heard a case of people using apt to do something similar (I've no idea if yum is up to the job yet though).
I've used an HL1440 and HL1450 and they do seem to be reliable and work well with Linux (the only thing I wish they supported is postscript but it's no big deal and they do PCL emulation).
HP's Laserjet 4 is what I would pick up second hand. Those things are like tanks and I'm sure they last for decades. Stay away from HP's Laserjet 5 (you feed the sheet in the top) as the one's I've seen always wind up being nothing but trouble after a few months (stop picking up paper, jamming etc). I'm also less impressed by the Laserjet 4000 series printers (they just don't seem to be as sturdy as the 4).
Apple Laserwriters also seem good in the second hand market.
A 2.6 distro will probably feel quite snappy on modern (1Ghz+) machines (RH9 is still 2.4 based with a few patches to reduce latency but there is no kernel preemption and that makes a noticible difference).
Robert Love pointed out that BeOS often traded throughput for latency (point 2) coupled with that fact it was highly threaded for its reknowed responsiveness. On weaker (PII or less) machines I suspect BeOS will always feel faster but there will come a point where CPUs context switch so quickly extra effort beyond preemption will not be worth the effort.
By the way, most desktop OSes these days have preemptive multitasking. It is only older OSes such as OS9 and Windows 3.11 which do not.
Some organisations that receive spam reports either have wrong contact information or ignore the spam reports. If spamcop sends lots of reports and they bounce (bad email address) or are ignored (e.g. the ISP doesn't like receiving (munged) spamcop reports and has told spamcop not contact it or the ISP never does anything with them) then the report goes to/dev/null but a record is kept so that they can be statistically tracked.
The Economist's The Internet Security story on slashdot earlier talked about how the gap between the fix for an exploit being announced and the exploit being used in the wild has be shrinking on Windows:
Attacks are also happening faster. A few years ago, it typically took virus writers a year to exploit a software vulnerability announced by a vendor. This gap between disclosure of a flaw and attack has been shrinking. For the Slammer worm in January it was six months, and for Blaster in August a mere three weeks. It is almost three weeks now since Microsoft brought out its patch for Windows 2000.
The sad thing about this case is that the way the story has broken before all the parties had fixed up kernels out (I think we're still waiting on RedHat dunno about SuSE). I kind of wish they had held out another day but maybe they had waited long enough already. Kernel exploits are nothing new but they sure are unpleasent.
gcc and glibc aren't controlled sorely by redhat but those are the things that tend to break binary compatibility. So if the (upstream) gcc / glibc folks break compatibility but say povide a 50% speed increase or improved security then Fedora will probably break binary compatibilty too. Just for the record a fair few (redhat 8/9) rpms continued to work in fedora but some did break and needed recompiling (e.g. pine was broken by updated kerberos libraries). Other times you need to install a compat library and things compiled with gcc 2.x will generally run into trouble.
In the kernel there was bearly any binary compatibity to start with... This is mostly an non-issue if you have the source but obviously if you do not want to / cannot recompile the source for most of what you use you are better off with something else that has a longer support cycle.
Thanks for following up on this. Your other post went on to describe why signatures haven't caught on.
On the later point, I can't help thinking that every little helps - take firewalls for example. You can argue that you shouldn't need a firewall if you keep your software up to date and configure the softwareto only listen and respond to appropriate network requests. Having a firewall as an extra layer may help cover or detect a slip up elsewherem helping to cut back on a particular type of incident. Signatures on packages could help detect intrusions such as the one that occured and help automate part of the verification step.
Another argument for signatures could be that the debian servers are simply bigger targets than a debian developer's box. As such they will draw more fire and surely the extra time spent watching the server may be better spent watching developers...
Are deb's signed? (I'm not that familiar with debian but I'd imagine they are) If so then just tell apt-get to not install debs that don't match a known signature...
Slashdot Mirror
Those who have done tests have found that kernel preemption reduces average latencies rather than worst case latencies (e.g. when your audio drops out or window stutters you are seeing worst case latencies) at the cost of hurting throughput a tiny bit (so things actually take longer because they are interrupted more). In general, the low latency patches had a bigger effect than preemption which is why many of the distros do not ship with a kernel with preemption on by default. However it is useful as a debugging aid and may well be a stepping stone to something else.
I suspect you will find more of the cheats that Window's uses making their way into Linux desktop distros because the user perception is improved. Currently I'm willing to trade a bit of RAM for a faster startup but it's how to do that so I do not notice that you are doing so is the real trick.
I don't know if it's possible for all the icons (e.g. those in the panel, open office etc.) but for window decorations and nautilus just change the GNOME theme from the GNOME control panel to something like Traditional.
And quite buggy. Not quite sure what that's got to do with Virtual PC though.
Taking a look at the torrent file seems to suggest that the DVD version is just an ISO and nothing else. Be careful folks - without *signed* MD5SUMs you can't be sure of what you've downloaded.
Apparently there are some systems that yum simply can't handle because it has to update the system while it is "online" (e.g. LVM). So it looks like the answer is "it depends on your set up".
See Seth Videl's post about it. My advice is to wait and see what the pitfalls are since there *will* be gotchas.
...hmm. advogato's being a bit strange today so let me post a quote:
My understanding is that the *Universities of Kent and Lancaster* are no longer providing a mirroring service. There will be some sort of mirroring service provided to JANET users by a different third party. I think there was a tendering process and JANET decided to go with someone else this year.
However, I'm not sure that it's clear whether the new mirror will:
The last point is the real sticking issue. Can anyone else clarify things? Either way mirror.ac.uk (as it is currently) will be sorely missed. It's provided an extremely useful service over the years and I'm sure it's saved Swansea Uni a lot of transatlantic traffic over the years :)
Read this article on the hoped move to Users with least priviledge in Windows Longhorn - it's just too risky to leave people running with maximum power all the time.
You're right preempt is turned off in Mandrake's stock kernels but it doesn't make a difference in this case since preempt only affects average case latencies not the worst case ones in a 2.6 kernel according to benchmarks.
The one big problem I've found with sophos is that the scanning engine itself will only work for 3 months before it has to be updated. This alone is incredibly frustrating.
You probably mean supermount as opposed to automount. You will be pleased to hear that supermount is no longer used by default and the Redhat / Fedora way of mounting is now used. A quick word of warning though - this has been quite a buggy release...
As to the autorun thing - were you using an NT/2K/XP machine? I'm fairly sure it does what is supposed to on 9x machines where you are more at liberty to poke memory but I'm fairly sure this process failed on NT series.
Just grab CD 1 and do a minimal install. Assuming you know the nams of the packages you want, you can (after configuring it) use yum to network install the rest once your system is up.
I'm fairly sure it is possible to do such minimal installs with Mandrake too (I've never tried) - just make sure you install only the minimum and be prepared for more hastle.
Have you tried contacting them? Email webmaster [at] mandrakeclub [dot] com and it's more like to be resolved. Seriously don't hang around on this (I know from experience). If you just complain on Slashdot your account may never be activated...
So what do you think happens when that proxy needs to make to make a link to a connection beyond the confines of ja.net because the latest page you are fetching from Slashdot has changed? It sends a connection out through the ja.net network and off via the transatlantic link. ja.net then record the amount of data sent across it and bill your University appropriately. Just because the immediate connection from the university is a ja.net link does not mean your uni won't be charged (per the Mbit) once it crosses out of ja.net.
I think you will find that UK Universities DO pay bandwidth costs for traffic that goes outside of ja.net (maybe they are charged for ja.net traffic too these days). Prices are especially steep if you generate transatlantic traffic. Caching of data has become very important in an effort to cut costs and it's rare that you are allowed to make www connections without going through some sort of Uni proxy.
Take a look at the network charging page for more details.
Why on earth would the bandwidth be free? Just because it's "academic"?
Well the problem is that not everyone starts out using MS software. You have already pointed out Emacs (which might predate Windows but do you have any idea how frustrating it is to press ctrl-w to delete a word and have it close a window instead?) but the fact of the matter is something like Mozilla's newsreader may have inherited it's keybindings from Netscape's newsreader. Those defaults simply won't change because those were the ones that were there first. If I'd been using an app for years the last thing I want is for the keybindings to change to something else just because the developer wanted to be compatible with the competition so clearly that's not what is going to happen.
Perhaps you are aware that Word's keybindings are not indentical to those of Word Perfect's (despite the later having a massive market share when Word initially came out). The trick is to provide a migration mode initially to help people settle in. Perhaps those doing the complaining could implement a "MS Word User" option in OpenOffice that binds keyboard settings to those of Word. In the end users will give up the bindings and switch to the new one but it would help the inital process. I believe MS may have done something similar to help Netscape users switch from NS3 to IE4...
... with support on a Redhat system it appears wise to pay for Enterprise Linux that will be supported for at least 5 years (http://www.redhat.com/software/rhelorfedora/). If you want to avoid a regular upgrade cycle with Redhat then you have to be prepared to pay (which seems fair to me since backporting costs increasingly more as time goes by). Whether five years is enough is another question altogether though...
You are absolutely right about the old RPM db being prone to becoming corrupt in releases earlier than Fedora Core 1 and more than a few people have been burned by this. After Alan ran into such a corruption on one the Redhat 9 + ximain machines he was updating on SUCS he mentioned that such things should be a thing of the past from Fedora Core 1 onwards as RPM had been made transactional...
By the by, I have seen redhat be updated from 7.2 to Fedora Core 1 (stopping by 8.0, and 9) using up2date over a network connection (I don't know if this is approved either but it's all much of a muchness now unless you are using WS or ES). I think I have also heard a case of people using apt to do something similar (I've no idea if yum is up to the job yet though).
I've used an HL1440 and HL1450 and they do seem to be reliable and work well with Linux (the only thing I wish they supported is postscript but it's no big deal and they do PCL emulation).
HP's Laserjet 4 is what I would pick up second hand. Those things are like tanks and I'm sure they last for decades. Stay away from HP's Laserjet 5 (you feed the sheet in the top) as the one's I've seen always wind up being nothing but trouble after a few months (stop picking up paper, jamming etc). I'm also less impressed by the Laserjet 4000 series printers (they just don't seem to be as sturdy as the 4).
Apple Laserwriters also seem good in the second hand market.
A 2.6 distro will probably feel quite snappy on modern (1Ghz+) machines (RH9 is still 2.4 based with a few patches to reduce latency but there is no kernel preemption and that makes a noticible difference).
Robert Love pointed out that BeOS often traded throughput for latency (point 2) coupled with that fact it was highly threaded for its reknowed responsiveness. On weaker (PII or less) machines I suspect BeOS will always feel faster but there will come a point where CPUs context switch so quickly extra effort beyond preemption will not be worth the effort.
By the way, most desktop OSes these days have preemptive multitasking. It is only older OSes such as OS9 and Windows 3.11 which do not.
Some organisations that receive spam reports either have wrong contact information or ignore the spam reports. If spamcop sends lots of reports and they bounce (bad email address) or are ignored (e.g. the ISP doesn't like receiving (munged) spamcop reports and has told spamcop not contact it or the ISP never does anything with them) then the report goes to /dev/null but a record is kept so that they can be statistically tracked.
The sad thing about this case is that the way the story has broken before all the parties had fixed up kernels out (I think we're still waiting on RedHat dunno about SuSE). I kind of wish they had held out another day but maybe they had waited long enough already. Kernel exploits are nothing new but they sure are unpleasent.
gcc and glibc aren't controlled sorely by redhat but those are the things that tend to break binary compatibility. So if the (upstream) gcc / glibc folks break compatibility but say povide a 50% speed increase or improved security then Fedora will probably break binary compatibilty too. Just for the record a fair few (redhat 8/9) rpms continued to work in fedora but some did break and needed recompiling (e.g. pine was broken by updated kerberos libraries). Other times you need to install a compat library and things compiled with gcc 2.x will generally run into trouble.
In the kernel there was bearly any binary compatibity to start with... This is mostly an non-issue if you have the source but obviously if you do not want to / cannot recompile the source for most of what you use you are better off with something else that has a longer support cycle.
In January, Jamie Zawinski made a post about the possibility of running a debugger backwards...
Thanks for following up on this. Your other post went on to describe why signatures haven't caught on.
On the later point, I can't help thinking that every little helps - take firewalls for example. You can argue that you shouldn't need a firewall if you keep your software up to date and configure the softwareto only listen and respond to appropriate network requests. Having a firewall as an extra layer may help cover or detect a slip up elsewherem helping to cut back on a particular type of incident. Signatures on packages could help detect intrusions such as the one that occured and help automate part of the verification step.
Another argument for signatures could be that the debian servers are simply bigger targets than a debian developer's box. As such they will draw more fire and surely the extra time spent watching the server may be better spent watching developers...
Are deb's signed? (I'm not that familiar with debian but I'd imagine they are) If so then just tell apt-get to not install debs that don't match a known signature...