October 26, 2026 at 4:16 PM, "Alperen Erkan" <[email protected] 
mailto:[email protected]?to=%22Alperen%20Erkan%22%20%3Cerkanalperen54%40gmail.com%3E
 > wrote:



> 
> Hello
> 
> > I follow OpenBSD development for fun, and there is an interesting new 
> > filesystem proposal:
> > good enough fs (gefs), which is a "crash-safe, snapshotting, copy on write 
> > FS".  It might
> > become the default filesystem on Plan9.  The developer uses his gefs to 
> > create a snapshot
> > every 5 mintues, which seems pretty cool.
> 
> Interesting.
> 
> > Apparently it's 9,000 lines of C code in the kernel.  The license is MIT.  
> >Maybe a port to
> > the Hurd would be worthwhile ?
> 
> Before considering a possible port to GNU Hurd, I would like to raise a few 
> technical questions:
> 
> Integration with Hurd: Would implementing GEFS as a userspace filesystem 
> server using Hurd's existing interfaces be the most practical approach? Are 
> there any architectural limitations we should consider?
> 
> Bε-tree performance: Since GEFS relies on buffered tree operations, how might 
> Hurd's IPC overhead affect its performance, particularly under 
> metadata-intensive workloads?
> 
> Crash consistency: What guarantees would the underlying storage interface 
> need to provide to preserve GEFS's copy-on-write consistency model across 
> crashes?
> 
> Snapshot management: Are there any known limitations in GEFS's 
> generation-based block reclamation and deadlist mechanism that could 
> complicate its integration with Hurd?
> 
> Portability: Would it be more reasonable to isolate GEFS's storage engine 
> from its Plan 9-specific interfaces before attempting a Hurd port?
> 
> I think these questions could help determine whether GEFS is a suitable 
> candidate for an experimental Hurd filesystem.
> 
> I have many more questions, but I wanted to start by asking these 5 
> fundamental ones.
> 
> I'd be interested in hearing your thoughts.
> 
> Best regards,
> Alperen Erkan

I'm just forwarding your email to bug-hurd.  Let's keep the discussion on the 
mailing list eh?

And I really do not have any answers to your questions.  I'm not a Hurd 
developer.  :(

Maybe is some ways ext3/4 is better that gefs.  Though I don't think ext3/4 has 
snapshots...

Joshua

> On Mon, Oct 26, 2026 at 9:56 PM jbranso--- via Bug reports for the GNU Hurd 
> <[email protected] mailto:[email protected] > wrote:
> 
> > 
> > Hey Hurd people!
> > 
> > I follow OpenBSD development for fun, and there is an interesting new 
> > filesystem proposal:
> > good enough fs (gefs), which is a "crash-safe, snapshotting, copy on write 
> > FS".  It might
> > become the default filesystem on Plan9.  The developer uses his gefs to 
> > create a snapshot
> > every 5 mintues, which seems pretty cool.
> > 
> > Apparently it's 9,000 lines of C code in the kernel.  The license is MIT.  
> > Maybe a port to
> > the Hurd would be worthwhile ?
> > 
> > https://undeadly.org/cgi?action=article;sid=20260918115729
> > 
> > the gefs paper is here:  https://orib.dev/gefs.pdf
> > 
> > Thanks,
> > 
> > Joshua
> >
>

Reply via email to