-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Tuesday 2008-07-22 at 15:58 -0700, Linda Walsh wrote: ...
Also -- I was wondering what people (if anyone) is thinking about doing for compatible filesystems for USB-thumb/flash drives?
Wasn't there a 4G limit on FAT32?...or is that just XP creation?
Yes, that limit exists. There is an article in the wikipedia that lists all the limits and who imposses them.
Still with a 32-bit FAT (1G), isn't it pretty much the case that the FAT's themselves need to be resident in memory all at once to maintain consistency? That sorta limits how big volumes might get.
The volume size is quite big, much bigger than what the windows formatter allows. Again, the wikipedia explains it.
With a 32-bit FAT, all full, that would take what -- (assuming signed 32-bit integers, that's only 2-G blocks. I'm not sure what size blocks people are willing to go with -- but assuming an 8-K block size, we're still limiting volume sizes to 16-terabytes -- WITH a 2G FAT table -- which would be huge -- if not entirely impractical in practice. Just *guessing*, but anything more than maybe 8-16M of FAT table would be stretching FAT, using that as a limiting factor, that constraings FAT volumes to 8K(8M-16M)=64G-128G, which is still 3-4 years away for mem drives (8G chips were available for cameras a year ago...so doubling every every year would be 64G in 2, every 18 months=3 years.
The filesystem size limit is the number of clusters and the size of each. Fat entries (number of clusters) are 268,435,437 (2^8-19) for fat32.
But main prob with FAT32 is the 2G file-size limitation.
4GiB, not 2.
I started hitting that 2-4 years ago, at least. The other problems -- that large FAT size that ideally needs to fit in memory, might be redone to allow partial mappings -- but I think the device would need to be locked while a process is updating the FAT map -- certainly not ideal.
It no longer needs to fit in memory. Locking of the FAT is no problem, at least on linux, because the kernel is the only one doing the writing.
I know there are plenty of file systems on linux -- but virtually none of them are ported to Win32, and I can't see NTFS becoming a defacto-industry standard as long as MS sits on it as proprietary.
There is ext2/3 for windows.
Already, on newer electronics w/cheap ROM's (phones), some are limited to using 4G chips because of addressability problems. Perhaps ideally, such a file system would get input from flash-ram designers to optimize file system operations with what's optimal for HW now -- and is likely to be optimal in the future.
They could use ext2/3 if they wanted to, or others, as many manufacturers of embedded machines do. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFIhm3NtTMYHG2NR9URAs4jAJ9DHe98IBr+/BmcP6Gzx7c6xXiVWACgki+v rtfw4v8nIJJiSwfMxZ5C3Ro= =+fex -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org