Guarantee Critical Regions in Portable code. Portable Semaphores?

The file system in ISIS II for Intellec development system used a multilevel linked list structure.

Paul

But prior art does count, so even though the USPTO issues a patenet doesn't necessarily make the patent defendable. It's left as an exercise to the patent holder to defend and any interested civilian to contest an issued patent if there's a question about its validity. This leads to situation where it's cheaper for the "infringer" to pay some nominal amount to the holder of an invalid patent than it is to contest the validity. There are less than scrupulous companies that will infringe patents knowing the patent holder can't afford to defend it.

Yep, and there are less than scrupulous companies that hold patents that they know are invalid, but they can out finance anyone that tries to take them on. I have one such customer. In the words of Glenn Campbell in "True Grit", they pull their lawyer like a gun.

But was the linked list stored in a table or threaded through the blocks?

AppleDOS used linked lists threaded through the blocks - the directory entry (there was only one directory) pointed to the first block.

ProDOS had a more efficient scheme, but I don't remember the details offhand.

George

The law is such that there is the presumption that all due diligence was done before the patent was granted and so the patent is presumed legitimate. That gives the patent holder the high ground in any dispute. It can be incredibly difficult and costly to overturn a patent even if you have undisputable proof of prior art.

George

As I understand it from talking with (the highest priced partnership firm in my large city's area) patent attorneys about this, even that can depend on the jurisdiction. Other than that, your comments conform to what I remember -- they specifically mentioned what you wrote, namely that a patent holder often has the high ground in a dispute. It was just that I remember them immediately countering their own comment, stating that "Some jurisdictions give the patent holder NO advantage."

Jon

My father and sister are patent attorneys (my father is a partner in a national firm, my sister works for Kodak (at least for now - she appears to have survived the latest round of downsizing)).

IANAL.

The presumption that the patent is good is written into federal law. The particular court matters mostly in the willingness of judges to issue injunctions against a suspected infringer - which only benefits a patent holder (it's impossible to get an injunction against a patent holder without first overturning the patent). Certain places, like Pennsylvania and Texas, are known for easily granting injunctions. In other places, like New York and Virginia, judges tend to be more cautious about issuing them. I think that's probably what your lawyers were referring to.

IANAL George

Thanks for the clarification. That makes sense within the context I remember, from more than 20 years ago.

Jon

On Jan 31, 5:44=A0pm, CBFalconer wrote: [snip]

IIRC patents are for 17 years. (Why the odd number? I don't know.) Ed

my guess: Used to be 20 years from data of filing, now 17 from granting. From filing to grant can take up to 3 years. It changed when the international patent treaty became effective, so international issues probably played a role.

In message , FreeRTOS.org writes

There USED TO BE a license for MS LFN on their web site. 2005-6 but I can't find it now. You had to apply and there was a 5 cent (I think) royalty charge

As far as I know MS only sued 2 companies a long time ago for using LFN (and I believe only claimed for the legal coasts) to set the precedence that they were protecting their IP.

Not water tight but it is highly unlikely that they will sue YOU unless

1 you are a direct and major competitive threat 2 you have lots of money (to make it worth suing you) 3 you have damaged MS in some other way and they are just adding this on top

--
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
\/\/\/\/\ Chris Hills  Staffs  England     /\/\/\/\/
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/

In message , Vladimir Vassilevsky writes

And most if not all OS have their own file systems either as part of the OS or a plug in.

So anyone with an OS will almost certainly get the file system from the same source.

--
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
\/\/\/\/\ Chris Hills  Staffs  England     /\/\/\/\/
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/

In message , james.walmsley writes

Correct but... you need to test on something like 50+ platforms?

Trouble is portable can be a lot less efficient. Most systems have a split between the portable and target specific parts.

It is bad because they system may not be safe if the interrupts are disabled. (This is the embedded NG? )

Define "long period of time" that works for all systems that your code will be ported to. :-)))

Define "very short period of time" that works for all systems that your code will be ported to :-))))

Hence the naive comment from the other poser ....

--
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
\/\/\/\/\ Chris Hills  Staffs  England     /\/\/\/\/
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/

Other way around.

Disabling interrupts briefly is a very effective way of implementing semaphores. Many real time operating systems do exactly that. This can be used to safely implement your own semaphore system - on a single processor/core system.

Disabling interrupts is sort of like not using gotos. Some people will say never use them even under threat of torture, but they're still very valid tools when used properly.

And if you're writing your own OS, or your OS doesn't provide this, or your OS is buggy, or you don't even have an OS?

Basically you have to use the facilities provided by your hardware, if the OS doesn't do this correctly for you. A great many real time operating systems do not provide cross processor atomicity.

I'm not guessing that this FAT library is being designed for very beefy systems, or that it's supposed to be extremely general purpose for every application and use. If there are some systems where disabling interrupts for a few instructions at the wrong time is disasterous, then this library wouldn't be appropriate on that system. Anyone trying to port that FAT library to their system should know what these pitfalls are.

d

Yes, this is true. And the OS versions will inevitably be a lot more efficient (the kernel could put the waiting task to sleep, will know about process priorities, etc). This could be a bit more work when porting, especially if the OS doesn't have semaphores though, so maybe provide sample implementation?

Now the whole bag of worms about priority inversion and deadlock and the like arises :-)

You can have no OS, but still have threads. Depending upon how you define "OS" :-)

Op Mon, 02 Feb 2009 18:43:33 +0100 schreef james.walmsley :

If and only if there will be no FS calls from an interrupt context (which is not uncommon for event-driven systems).

--
Gemaakt met Opera's revolutionaire e-mailprogramma:  
http://www.opera.com/mail/

Just as an extra note to this thread:

FullFAT is now "complete", currently about to release RC4. Version 1.0 will be available in about 1 month's time.

See

for more information.

FullFAT is available from

For those interested in how FullFAT works regarding the issues raised i this thread, it creates a platform portability layer, allowing FullFAT t call OS provided Semaphore services. Or a developer can use another simila mechanism for his/her platform.

James