[OpenAFS] OpenAFS vs NFSv4?

Ryan Underwood nemesis-lists@icequake.net
Mon, 28 Apr 2003 12:22:28 -0500


On Mon, Apr 28, 2003 at 09:46:16AM -0400, Ted Anderson wrote:
> On 4/28/2003 08:02, Nathan Neulinger wrote:
> >Support for a reasonable number of platforms.
> >Location transparency.
> >Transparent on-the-fly moves of data.
> >Single namespace automatically. 
> >Built in snapshotting mechanism.
> 
> To Nathan's list of AFS advantages, I'd add:
>   + Directory caching (NFSv4's delegation doesn't work for dirs)
>   + Volumes
>   + Simpler, per-directory ACLs
> To be fair we should also add AFS negatives:
>   - Proprietary data format (NFS exports regular file systems)
>   - Large file support
>   - Byte range and session locking
> And NFSv4 plusses and minuses:
>   + Kerberos 5 and LIPKEY authentication
>   + Stronger stream encryption (DES vs FCrypt)
>   + Complex, but Windows compatible, per-file ACLs
>   - Many useful features are optional
>   - Immature: Available in alpha-quality implementations only
> 
> I'm sure there are many more items that could be added to each list.
> 
> I would also emphasize the point that Daniel Clark raised about NFSv4: 
> many of the interesting features that move it towards AFS are optional 
> (e.g. delegation, but not strong authentication).  So, as NFSv4 is 
> deployed it will be important to see which features really become 
> wide-spread.

Thanks for all the input; it is much appreciated.

-- 
Ryan Underwood, <nemesis at icequake.net>, icq=10317253