Hi Kiste,

2011/5/6 Oliver Seitz <[email protected]>

>
> *Matt, I read in comments your FAT32 lib doesn't support fragmented files.
> Do you actually try to implement this feature, and did you give up because
> not enough resource ?**
> *
>
> I feel adressed as "guy" and try to answer as much as i know ;-)
>

Great !


>
> There are two approaches to deal with fragmented files:
>
> 1.) Start reading, and as soon as we reach the end of the fragment, go back
> to the FAT and find the next fragment. This could need a lot of time
> occasionally while reading in the middle of a file. The same FAT sector may
> have to be read multiple times during acces of a single file.
>
> 2.) When opening, parse the FAT and collect all the fragments to a buffer.
> Needs only time when opening the file, but needs a buffer that could limit
> the allowable number of fragments.
>
> Matt chose to use this second variant, which IMO is a good choice
> generally. It is not universal, however. Theoretically, a file of 2GB can
> consist of four millions of fragments. The 16MByte (or more?) buffer that
> would be needed is beyond anything you could reasonably connect to a PIC.
>

OK, back on Minix world, I chose number 1: I explore zone (ie. block)
addresses from inode itself and indirectlyn while reading and writing. I
understand better why fragmented files aren't supported by fat32 current
impl. I could indeed do the same in a restricted version of Minix. Zone
exploration is done in minix_next_zone(). I could write another simpler
version which would just return address + 1. I would still need to have
proper inode content to make readable on another system, so I wonder how
"heavy" this version would still remain... I'll see what I do what I'm done
with most importannt feature implemented (eg. I'm not going to support
symlinks I think)


>
> This is the reason that made me think about a "new simple" file system that
> uses a kind of linked list. By mixing user data and meta data in each
> sector, there would be no need to neither cache fragment data nor jumping to
> a system area to find the next sector.
>

I've no doubt about how useful such a "simple" filesystem would be ! And
thus be happy to give a try when possible.

Cheers,
Seb

-- 
You received this message because you are subscribed to the Google Groups 
"jallib" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/jallib?hl=en.

Reply via email to