Re: [PATCH]tar filesystem for 2.6.10-rc1-mm3(easily access tar file)

[Tomasz Kłoczko]

On Sun, 7 Nov 2004, andyliu wrote:
[..]
> > > Simply wonderful!
> >
> > Which is ~equal to .. unpack tarfile.tar to /dir/to/mnt :o)
> if the tarfile.tar contain huge little file and unpack it will cost
> time and much disk space.

1) This tarfs code have only RO support. Prepare RW acces for tar nball
   file structure will not be so trivial.
2) If You want save disk space for small files with also RO access
   only better will be spend some time on romfs for extend them for allow
   operate on slightly bigger file systeme than curreent lmitation this fs.
   Packing to tar file will be functionaly equal to generate romfs image
   using genromfs.
3) Maybe I'm wrong but IIRC raiserfs have block suballocation. If I'm
   wrong (about current raiserfs) btter will be write and use some fs with
   this kind feacture.
4) Huge or very huge amout of small files it is usulaly case on filesystem
   for mail spool. It need RW access. Subject was many tilme in past
   discussed and can be find on archives aroud mailfs or mailstorfs
   words. Tar ball structure isn't specialized for allow fast RW operation.
   In cases where speed isn't issue tarfs will be fuctional equivalet to romfs.

Count of cases where tarfs will be useable will be probably very close to 
cout of cases where romfs is used now. All without add single line of code 
on kernel level :-)

tarfs it is *good* code example but IMO .. only example :_)

kloczek
--
-----------------------------------------------------------
*Ludzie nie mają problemów, tylko sobie sami je stwarzają*
-----------------------------------------------------------
Tomasz Kłoczko, sys adm @zie.pg.gda.pl|*e-mail: kloczek@xxxxxxxxxxxxxxxxxx*