Re: New Linux 2.5 - 2.6 TODO (Alan Cox suggestsdelaying

[James Sutherland]

On Mon, 5 Jun 2000, Hans Reiser wrote:
> Alan Cox wrote:
> > 
> > > ;-) I don't think either FS will get through 2.5 without a major logging
> > > overhaul.  I've got some hash tables I would love to throw out, and
> > > will probably integrate all the logging into the page cache.
> > 
> > This basically is what all of this comes down to
> > 
> > Do we
> > 
> > A.      Add all these various journalling file systems in 2.4 and try and
> >         sort out the common journalling/support code in 2.5
> > 
> > B.      Defer merging with the main tree on the assumption all the vendors
> >         can and will do merging so that when we put them in the main tree
> >         we do so with the common stuff sorted out properly first
> > 
> > And thats a call Linus has to make not me
> 
> I suggest that we add ReiserFS in 2.4 because it works

Nothing to do with the 200% revenue increase it will give you? It seemed
ironic, seeing the complaint by you about Red Hat and SuSE trying to
manipulate the kernel to their own commercial ends...

> and 18 months is a long time to go without journaling,

Who will be going without journalling? If anyone wants/needs ReiserFS
support, they are free to patch it in. I'd be rather reluctant to, since
the only ReiserFS user I know personally lost the entire partition...

> rewrite journaling to implement wandering logs in 2.5,

How is this relevant to the 2.4 discussion?

> and then consider merging in 2.7.  2.5 is way too early for merging.  

You want to merge the current codebase into 2.4, but you claim 2.5 is too
early for a merge?

> Keep in mind that wandering logs are probably going to be implemented
> by Zam not Chris, as Chris is more interested in other tasks, and it
> integrates nicely with Zam's allocate on flush coding.

Again, how is this relevant? You have a chunk of code you want included in
the kernel during a code freeze. Either it does get included, or it
doesn't. The improvements planned for the 2.7 kernel - which could be 4
years away - aren't really relevant to this.

James.