a major weakness of his argument is that he implies (credibly) that this lab test commissioned by Apple is not trustworthy, and then compares it with tests by Dell and Intel, which he seems to present as implicitly true
He only mentions the results that Dell and Intel produced for Dell and Intel platforms (all the Apple results are the ones given by Apple). In other words, if Apple wants to be biased in favor of G5, so should Dell/Intel and that is exactlly what he presents. It sounds like a valid argument to me!
I don't think that by changing permissions on the file you can prevent someone from *mailing* or *copying* the file to a new location and changing it. That's what companies really want, global access control, not access control for each copy of the document.
Medical companies need to record every detail on the experiments they do. Result data should definitely be read-only, no matter who copies them where. I don't think permissions can cover that.
I am not a MacOSX user, I am thinking of investing in a Powerbook though. What worries me is that there seem to be more than one ways to do certain things:
Symbolic Links: You can make them through the GUI, but Darwin can't use them, or you can run ln -s and Darwin+Cocoa+Carbon will see them
Hidden files: you can hide files using the ".", but you can also hide them using another file
Something else?
Any insights by UNIX geeks that use MacOSX? Have you been bitten by these "features"?
everyone forgot about Xt which works beautifully, and decided to make their own widget sets. this is really annoying when trying to embed Xt stuff into applications that use gtk or qt.
Xt was (is) just a toolkit framework on top of X, it does not change or modify the X protocol. Not only that, but Xt is a mediocre attempt at a toolkit, compared to modern standards: programming with Xt is not easy or intuitive and the on-screen widgets are not up to it.
Xt is not the answer, but a unified toolkit would be nice. I don't think it will happen though, not in this lifetime.
Just because some people exploit them doesn't mean they aren't relevent. They are still an important ingredient of HTML soup.
I will have to agree with the parent.Metadata, in general, are necessary if you want to figure out what your data is all about. Unless you have self-describing data (which is something that XML promises to do), you have to deal with metadata as well. So, until everybody publishes their data in XML (and use XSLT to draw them on a browser), I believe that metatags have a value.
When Microsoft has undocumented, private, internal APIs, everyone cries "Foul!" and accuses them of hiding these APIs from developers. When they then change those internal APIs, everyone again cries "Foul!" and accuses them of breaking these internal APIs intentionally. But when Apple does this, it's okay? I guess I just don't get it.
I guess the argument here is that Microsoft uses these internal APIs to give their software an edge in the competition. When they change these APIs, they make sure to update their software in advance, so that again the rest of the world is lagging behind. Standard business practice, I guess...
While a trash can is nice to have, this doesn't fundementally address the issue of retrivability of accidently deleted information. That is, there is still going to be a step where information is going to be classed as unretrivable even when it COULD be retrieved. (i.e. when the trash is emptied)
Even with your solution, the file will be deleted eventually (when its blocks get reused). The way I see it, both solutions have the same outcome (the file becomes unrecoverable eventually) and yours requires requires non-trivial changes to the OS.
Besides, you can enforce some type of policy on your trashcan as well. You can assign, for example, priorities over the deleted files so non-important files are removed first from the bin.
The big advantage of libtrash is simplicity (it's a preloaded shared library - no recompiling or modifications required) and it's user space (no changes to the US).
> Forget transactions -- you can fake that with
> LOCKs.
Can you please explain how you will perform crash recovery using just LOCKs? Transactions are not just for concurrency control.
Condor http://www.cs.wisc.edu/condor uses
a checkpointing mechanism for migrating processes
between hosts (works on Unices & Win). Not
exactly what you need, but maybe a starting point.
Slashdot Mirror
a major weakness of his argument is that he implies (credibly) that this lab test commissioned by Apple is not trustworthy, and then compares it with tests by Dell and Intel, which he seems to present as implicitly true
He only mentions the results that Dell and Intel produced for Dell and Intel platforms (all the Apple results are the ones given by Apple). In other words, if Apple wants to be biased in favor of G5, so should Dell/Intel and that is exactlly what he presents. It sounds like a valid argument to me!
I don't think that by changing permissions on the file you can prevent someone from *mailing* or *copying* the file to a new location and changing it. That's what companies really want, global access control, not access control for each copy of the document. Medical companies need to record every detail on the experiments they do. Result data should definitely be read-only, no matter who copies them where. I don't think permissions can cover that.
Any insights by UNIX geeks that use MacOSX? Have you been bitten by these "features"?
systrace
everyone forgot about Xt which works beautifully, and decided to make their own widget sets. this is really annoying when trying to embed Xt stuff into applications that use gtk or qt.
Xt was (is) just a toolkit framework on top of X, it does not change or modify the X protocol. Not only that, but Xt is a mediocre attempt at a toolkit, compared to modern standards: programming with Xt is not easy or intuitive and the on-screen widgets are not up to it.
Xt is not the answer, but a unified toolkit would be nice. I don't think it will happen though, not in this lifetime.
I will have to agree with the parent.Metadata, in general, are necessary if you want to figure out what your data is all about. Unless you have self-describing data (which is something that XML promises to do), you have to deal with metadata as well. So, until everybody publishes their data in XML (and use XSLT to draw them on a browser), I believe that metatags have a value.
I guess the argument here is that Microsoft uses these internal APIs to give their software an edge in the competition. When they change these APIs, they make sure to update their software in advance, so that again the rest of the world is lagging behind. Standard business practice, I guess...
Even with your solution, the file will be deleted eventually (when its blocks get reused). The way I see it, both solutions have the same outcome (the file becomes unrecoverable eventually) and yours requires requires non-trivial changes to the OS.
Besides, you can enforce some type of policy on your trashcan as well. You can assign, for example, priorities over the deleted files so non-important files are removed first from the bin.
The big advantage of libtrash is simplicity (it's a preloaded shared library - no recompiling or modifications required) and it's user space (no changes to the US).
I am not sure if this is what you wish for, but Quanta looks very promising.
(Yeah yeah, it is a KDE app and all that but it still impressive)
> Forget transactions -- you can fake that with
> LOCKs.
Can you please explain how you will perform crash recovery using just LOCKs? Transactions are not just for concurrency control.
Condor http://www.cs.wisc.edu/condor uses a checkpointing mechanism for migrating processes between hosts (works on Unices & Win). Not exactly what you need, but maybe a starting point.