Re: [ros-dev] Security policy for FAT partition driver?

[Robert Köpferl]

>   
>  Well, in short, in the (now free downloadable) SFU 3.5 there is a NIS 
>  bridge, which exposes accounts and groups to NIS compliant machines. 
>  After installing SFU, your credentials management console shows a tab 
>  with tipical UNIX/POSIX attributes and files have an attrribute tab with 
>  rwxrvxrvx... It seems that the design of IFS subsystem assumes a 
>  superset of every known fs available about 1990.
>   
Hmmm, yea. That's basicly what I think of the wohle thing.
It should be possible for an IFS or another part to do a 1:1 or a 100:90 
  mapping of rights and users <-> SIDs

In my opinion there'S nothing to it to limit the user-part of SIDs to 
16-bit and cut out just that.


PS. My favorite is still JFS.
It has been prooved on OS/2 and OS/2 is our next relative. Its Src is 
open(GPL) and it supports most of the features. Even reparse points can 
be transparently implemented (EAs). I have nothing against s.o. 
implementing NTFS, but I see many problems. So cut & off

[ea]

Robert Köpferl wrote:

>   Hmmm, yea. That's basicly what I think of the
> wohle thing.
>  It should be possible for an IFS or another part to do a 1:1 or a 
>  100:90  mapping of rights and users <-> SIDs
>  
I guess this is done by the psxss.exe or by psxdll.dll, by mean of the 
LSA, which lookups user and group names from the LDAP db on the DC. No 
uid neither gid are stored in the fs metadata, only SIDs and ACLs, which 
are what the LSA can eat.

>   In my opinion there'S nothing to it to limit the
> user-part of SIDs to 
>  16-bit and cut out just that.
>  
Sorry not a POSIX guru here.

>  
>  PS. My favorite is still JFS.
>  It has been prooved on OS/2 and OS/2 is our next relative. Its Src is 
>  open(GPL) and it supports most of the features. Even reparse points 
>  can be transparently implemented (EAs). I have nothing against s.o. 
>  implementing NTFS, but I see many problems. So cut & off 

JFS is excellent on my experience in AIX: that OS has one of the most 
advanced and robust volume management I have seen.

[James Tabor]

ea wrote:
>   Robert Köpferl wrote:
>  
> >   Hmmm, yea. That's basicly what I think of the
> > wohle thing.
> >  It should be possible for an IFS or another part to do a 1:1 or a 
> >  100:90  mapping of rights and users <-> SIDs
> >  
>  I guess this is done by the psxss.exe or by psxdll.dll, by mean of the 
>  LSA, which lookups user and group names from the LDAP db on the DC. No 
>  uid neither gid are stored in the fs metadata, only SIDs and ACLs, which 
>  are what the LSA can eat.
>  
> >   In my opinion there'S nothing to it to limit
> > the user-part of SIDs to 
> >  16-bit and cut out just that.
> >  
>  Sorry not a POSIX guru here.
>  
> >  
> >  PS. My favorite is still JFS.
> >  It has been prooved on OS/2 and OS/2 is our next relative. Its Src is 
> >  open(GPL) and it supports most of the features. Even reparse points 
> >  can be transparently implemented (EAs). I have nothing against s.o. 
> >  implementing NTFS, but I see many problems. So cut & off 
>  
>  
>  
>  JFS is excellent on my experience in AIX: that OS has one of the most 
>  advanced and robust volume management I have seen. 
Found this;
http://jfs.sourceforge.net/


Thanks,
James

[James Tabor]

Robert Köpferl wrote:
>   
>  
>  PS. My favorite is still JFS.
>  It has been prooved on OS/2 and OS/2 is our next relative. Its Src is 
>  open(GPL) and it supports most of the features. Even reparse points can 
>  be transparently implemented (EAs). I have nothing against s.o. 
>  implementing NTFS, but I see many problems. So cut & off 
Wow! JFS! I see a good point here! IBM support for ROS! I did some digging
around and see many on the net would love to have a JFS-Driver for Windows.
This is a good idea!
James