I always thought it was a sick joke that Adobe Reader had so-called Bookmarks that were nothing of the sort. They were chapter and header markers.
I used to have the practice of writing a Stickie when I wanted to leave a note of where I stopped reading. Real bookmarks and the ability to annotate will prevent a lot of unnnecessary hassle and make PDF's almost as useful as they could potentially be.
My argument is that there could well be no performance hit whatsoever if the system would otherwise be twiddling its thumbs, waiting for the hard drive.
Multithreaded operating systems and filesystems do not "twiddle their thumbs". Unused CPU cycles are given back to the task scheduler which allocates them to processes waiting to be run.
The fact of the matter remains that with the filesystem encrypting the swap file, there is a new task required that is consuming CPU cycles that previously were made available to other tasks. CPU cycles are finite. As soon as you add a new process, it takes cycles away from other processes.
Frankly, I am more excited by Address Book sharing because Apple can take on Plaxo with something truly peer-to-peer and updating without annoying reminder emails. At least i hope so.
Multithreaded operating systems and filesystems do not "twiddle their thumbs". Unused CPU cycles are given back to the task scheduler which allocates them to processes waiting to be run.
If there are any. If they aren't also blocked on I/O.
I/O has this funny way of interrupting the normal course of things.
Quote:
The fact of the matter remains that with the filesystem encrypting the swap file, there is a new task required that is consuming CPU cycles that previously were made available to other tasks. CPU cycles are finite. As soon as you add a new process, it takes cycles away from other processes.
Which, as Kickaha points out, only matters in the very worst case, and then only if those processes aren't also blocking on I/O.
It remains obvious that, in all but pathological cases, VM encryption will be free for all intents and purposes. There are already implementations like this out in the wild, and nothing is getting bogged down. Your concern is at best theoretical. Unless Apple seriously drops the ball on the implementation, I don't imagine anyone will even notice the impact.
Multithreaded operating systems and filesystems do not "twiddle their thumbs". Unused CPU cycles are given back to the task scheduler which allocates them to processes waiting to be run.
Correction... Multithreaded OSs do not twiddle their thumbs if and only if they have something else to do. Realistically, almost ALL computers twiddle their copper and aluminum thumbs about 95% of the time... Only 90% for power users.
Oh yeah, only 5% for active servers. But the thing is, if you have an active server you care a lot more about security than a few CPU cycles. And, hopefully your server has enough memory that it doesn't have to write to or read from the page file very often.
In other words, the cost of this feature should be really low for just about everyone.
Correction... Multithreaded OSs do not twiddle their thumbs if and only if they have something else to do. Realistically, almost ALL computers twiddle their copper and aluminum thumbs about 95% of the time... Only 90% for power users.
Whatever - they twiddle 95% of the time if you include time the user is doing other things. If you are rendering scenes in some graphics/video app, its going to be slower with encryption- particularly because you ram is swapped out all the time.
Have no doubt that there is going to be a Mac out there that can just barely capture analog video-in without encryption, and can't keep up with it on.
However, I agree that for most people whom it will affect, they will just get a new machine. Its not going to be a serious issue.
Hey - did I just use "whom" right! If I did its the first time.
Does anyone know anything about the Dynamic DNS in the sharing tab? What services it works with? I'd like to try it out but I didn't attend WWDC and I don't have a build.
I'm with jhitchco. Tell us more about Dynamic DNS. Maybe someone at Apple saw my post from fifteen months ago where I said:
Quote:
Originally post by fahlman
I request that Apple adds a dynamic DNS application to 10.3, which in conjunction with your .mac account allows you to access your computer from any other MacOS X computer that has an internet connection and an Apple client application. With the growth of broadband connections with dynamic IP address this would not only make .mac a must have, but would also add one more feature to 10.3 that would make people want to upgrade.
Whatever - they twiddle 95% of the time if you include time the user is doing other things. If you are rendering scenes in some graphics/video app, its going to be slower with encryption- particularly because you ram is swapped out all the time.
The hard drive is going to be killing you there, not any encryption algorithm. Remember, a hard drive is three or four orders of magnitude slower to respond than a CPU, minimum. If your system thrashes when you do work, the solution is more RAM. There's no way around that.
Quote:
Have no doubt that there is going to be a Mac out there that can just barely capture analog video-in without encryption, and can't keep up with it on.
Given that every Mac CPU that can run video software at any usable speed can handle far more bandwidth than any external bus has, I sincerely doubt this. Again, sure it's theoretically possible, but in practice it's so pathologically unlikely that there's no point considering it. The hypothetical problems raised by the technology pale to nothingness beside the real benefits of the technology.
At any rate, if some machine exists for which this would be true, the problem would be solved the same way thrashing is solved: More RAM.
If there are any. If they aren't also blocked on I/O.
I/O has this funny way of interrupting the normal course of things.
Which, as Kickaha points out, only matters in the very worst case, and then only if those processes aren't also blocking on I/O.
It remains obvious that, in all but pathological cases, VM encryption will be free for all intents and purposes.
I have been experimenting with encrypted swap on my Mac for about a month now... I simply disabled the dynamic_pager in /etc/rc startup script and then wrote a little startup items script to create an encrypted Disk Image that holds the swap files.
I use a "sparse image", the same kind that FileVault uses, so that there is no overhead in disk space vs. unencrypted swap. I was only able to get this to work in Tiger (WWDC release so far). On 10.3.x sparse image for swap does not work.
Performance overhead? For me, NOT NOTICEABLE vs unencrypted swap.
Here is an encrypted-swap script by Andreas Schwartz (not me) that seems to work well with 10.3.x -- try it for yourself!
AlBook 15" 1.25 GHz, 1.25 GB RAM. When I'm swapping, I'm generally hammering the disk via application-level I/O anyway (Photoshop or Mail.app or compiling or copying large files to Firewire or running a large Java application (Eclipse IDE)).
Comments
Originally posted by macFanDave
Thank God for real bookmarks in Preview!
I always thought it was a sick joke that Adobe Reader had so-called Bookmarks that were nothing of the sort. They were chapter and header markers.
I used to have the practice of writing a Stickie when I wanted to leave a note of where I stopped reading. Real bookmarks and the ability to annotate will prevent a lot of unnnecessary hassle and make PDF's almost as useful as they could potentially be.
yes yes yes!
Originally posted by Amorph
My argument is that there could well be no performance hit whatsoever if the system would otherwise be twiddling its thumbs, waiting for the hard drive.
Multithreaded operating systems and filesystems do not "twiddle their thumbs". Unused CPU cycles are given back to the task scheduler which allocates them to processes waiting to be run.
The fact of the matter remains that with the filesystem encrypting the swap file, there is a new task required that is consuming CPU cycles that previously were made available to other tasks. CPU cycles are finite. As soon as you add a new process, it takes cycles away from other processes.
(Checks CPU meter)
Nope, not here. Sorry to hear your machine is that bogged down...
Originally posted by mikef
Multithreaded operating systems and filesystems do not "twiddle their thumbs". Unused CPU cycles are given back to the task scheduler which allocates them to processes waiting to be run.
If there are any. If they aren't also blocked on I/O.
I/O has this funny way of interrupting the normal course of things.
The fact of the matter remains that with the filesystem encrypting the swap file, there is a new task required that is consuming CPU cycles that previously were made available to other tasks. CPU cycles are finite. As soon as you add a new process, it takes cycles away from other processes.
Which, as Kickaha points out, only matters in the very worst case, and then only if those processes aren't also blocking on I/O.
It remains obvious that, in all but pathological cases, VM encryption will be free for all intents and purposes. There are already implementations like this out in the wild, and nothing is getting bogged down. Your concern is at best theoretical. Unless Apple seriously drops the ball on the implementation, I don't imagine anyone will even notice the impact.
Originally posted by mikef
Multithreaded operating systems and filesystems do not "twiddle their thumbs". Unused CPU cycles are given back to the task scheduler which allocates them to processes waiting to be run.
Correction... Multithreaded OSs do not twiddle their thumbs if and only if they have something else to do. Realistically, almost ALL computers twiddle their copper and aluminum thumbs about 95% of the time... Only 90% for power users.
Oh yeah, only 5% for active servers. But the thing is, if you have an active server you care a lot more about security than a few CPU cycles. And, hopefully your server has enough memory that it doesn't have to write to or read from the page file very often.
In other words, the cost of this feature should be really low for just about everyone.
Originally posted by HumanJHawkins
Correction... Multithreaded OSs do not twiddle their thumbs if and only if they have something else to do. Realistically, almost ALL computers twiddle their copper and aluminum thumbs about 95% of the time... Only 90% for power users.
Whatever - they twiddle 95% of the time if you include time the user is doing other things. If you are rendering scenes in some graphics/video app, its going to be slower with encryption- particularly because you ram is swapped out all the time.
Have no doubt that there is going to be a Mac out there that can just barely capture analog video-in without encryption, and can't keep up with it on.
However, I agree that for most people whom it will affect, they will just get a new machine. Its not going to be a serious issue.
Hey - did I just use "whom" right! If I did its the first time.
Originally posted by jhitchco
Does anyone know anything about the Dynamic DNS in the sharing tab? What services it works with? I'd like to try it out but I didn't attend WWDC and I don't have a build.
I'm with jhitchco. Tell us more about Dynamic DNS. Maybe someone at Apple saw my post from fifteen months ago where I said:
Originally post by fahlman
I request that Apple adds a dynamic DNS application to 10.3, which in conjunction with your .mac account allows you to access your computer from any other MacOS X computer that has an internet connection and an Apple client application. With the growth of broadband connections with dynamic IP address this would not only make .mac a must have, but would also add one more feature to 10.3 that would make people want to upgrade.
Originally posted by joshDavenport
Whatever - they twiddle 95% of the time if you include time the user is doing other things. If you are rendering scenes in some graphics/video app, its going to be slower with encryption- particularly because you ram is swapped out all the time.
The hard drive is going to be killing you there, not any encryption algorithm. Remember, a hard drive is three or four orders of magnitude slower to respond than a CPU, minimum. If your system thrashes when you do work, the solution is more RAM. There's no way around that.
Have no doubt that there is going to be a Mac out there that can just barely capture analog video-in without encryption, and can't keep up with it on.
Given that every Mac CPU that can run video software at any usable speed can handle far more bandwidth than any external bus has, I sincerely doubt this. Again, sure it's theoretically possible, but in practice it's so pathologically unlikely that there's no point considering it. The hypothetical problems raised by the technology pale to nothingness beside the real benefits of the technology.
At any rate, if some machine exists for which this would be true, the problem would be solved the same way thrashing is solved: More RAM.
Originally posted by averagezen
addressbook sharing is a huge feature for me and my wife with multiple computers in our home. ...
Hope they are also working on a way to share some Mail filtering rules...
Originally posted by Amorph
If there are any. If they aren't also blocked on I/O.
I/O has this funny way of interrupting the normal course of things.
Which, as Kickaha points out, only matters in the very worst case, and then only if those processes aren't also blocking on I/O.
It remains obvious that, in all but pathological cases, VM encryption will be free for all intents and purposes.
I have been experimenting with encrypted swap on my Mac for about a month now... I simply disabled the dynamic_pager in /etc/rc startup script and then wrote a little startup items script to create an encrypted Disk Image that holds the swap files.
I use a "sparse image", the same kind that FileVault uses, so that there is no overhead in disk space vs. unencrypted swap. I was only able to get this to work in Tiger (WWDC release so far). On 10.3.x sparse image for swap does not work.
Performance overhead? For me, NOT NOTICEABLE vs unencrypted swap.
Here is an encrypted-swap script by Andreas Schwartz (not me) that seems to work well with 10.3.x -- try it for yourself!
http://andreas-s.net/osx-encrypted-swap.html
AlBook 15" 1.25 GHz, 1.25 GB RAM. When I'm swapping, I'm generally hammering the disk via application-level I/O anyway (Photoshop or Mail.app or compiling or copying large files to Firewire or running a large Java application (Eclipse IDE)).