OSDev.org

Hi every one

I wanna know about filesystem and it's procedures.
I'm wondering that we can implement FS as a high-level service and/or module in OS.

Can it be proved?or is there any one who have a better idea?

****The hardest is not the best forever**** >>> ;D goofy

Administrator2004 wrote:
I wanna know about filesystem and it's procedures.

That's a bit of too high-level a question. Kind of like, "I want to know about philosophy"...

I'm wondering that we can implement FS as a high-level service and/or module in OS.

Every single OS project is doing so in some way or another.

Be a bit more precise. What do you want to know?

Microkernels traditionnally have their filesystem implementation in application-level services, if that was the meaning of the question ...

Even in a traditionnal architecture, if your process gets the R/W access on a partition and if you can install hooks in the kernel that call your process's functions, you're almost done.

The most complicated issue is probably for the "file server" to be able to commit or access memory pages within client address spaces... Also, i wouldn't let even the fileserver directly toy with disk controller registers nor with PCI dma programming ...

may be it is off the subject but i wanted to say something...

instead of discussing HOW to implement a filesystem, i think the time has come to think about new generation ( future ) filesystems like Microsoft's WinFS. ( dont freak out microsoft anticipants, WinFS is just an example, there are other types of database-like new generation filesystems developed in universities but they are not much known ) it is not the hard part how to implement it, but the hard part is actually to "design" it ( agree with me? )

so here comes my question:
-> what do you expect from a file "system".
-> how actually do we organize our for example home stuff. ( actually my stuff is all around my room but you know what i mean, there shouldn't be much difference between using a computer and doing ordinary day stuff, what are the computers are for anyway, not to ease our lives? )
-> is "file"s "link"s and "directory"s are "natural" ( may be i should have found a more precise word about that i mean by "natural" but i can't find the word now, i hope you understand me ).

ok.. a few brainstorming questions, just to think about.

Hi
thank you pypr.clicker and cemre but solar why do answer all questions like this?(please don't answer or if you want to answer please <<< say something that others benefit from not to distract others like me or...?""thank you""
and about you cemre::you have a good idea but it takes about half of hard disk to store these stuffs like : access times,last access,... .these are necessary but not in the way microsoft implemented because it will hurt old systems like mine.it's easy to do so!!!by implementing a batabase as a service in os and ofcourse you can access DB by C++ and/or php(easy) by using Zend API source files and GTK source files in os.
warning:;):without theory we can't code even an easy task switching needs a thoery before implementing.

and you Pype.Clicker:I mean something like what you said.


Last but not least::Thank you.

Disclaimer: Having a bad day today. On a better day I would have worded this in a friendlier way; the opinion remains the same though. Constructive criticism from my fellow moderators welcome.

----

Administrator2004 wrote:
solar why do answer all questions like this?(please don't answer or if you want to answer please <<< say something that others benefit from not to distract others like me or...?""thank you""

Dude, it's an attitude problem. (And it's not me having it, and yes I did take offense at your last post.)

"I want to know about filesystem and it's procedures." - that is not a question that can be answered. I could link documentation of <stdio.h> or <iostream>, on the AmigaOS dos.library, the Win32 file access API, ext2fs, ReiserFS, or other existing FS. I could link to papers on database-centric file systems, networking file systems, distributed file systems...

You know, I like artillery games, where you fire a shot somewhere, see where it lands, and then adjust direction, elevation, and power until you get close to the target.

In a forum like this, I prefer questions that allow people to get right to the point without firing a dozen tracer rounds to find out what you're about. Pype generally is more patient in this regard. My bad.

Besides, sometimes questions like yours give me the impression that the person asking the question didn't spend any significant time in doing research on prior art. Discussing at that level, I consider fruitless.

If that "distracts" you, too bad. Your questions distract me, because when I read the topic, I thought there might be someone having an actual problem, or wanting to discuss an actual design of his, not a general "let's talk about it but don't go into the details" type of thread.

Sorry but that's my opinion.