Re: Why no work on rsync?
Re: Why no work on rsync?
- Subject: Re: Why no work on rsync?
- From: Rob Braun <email@hidden>
- Date: Sat, 04 Mar 2006 12:18:24 -0800
Jordan K. Hubbard wrote:
On Mar 4, 2006, at 5:18 AM, Peter Seebach wrote:
I have a hard time imagining anyone but Apple ever caring about resource
forks. The inconsistency with the fundamental UNIX file model is simply
too great.
I'm not sure that's entirely accurate. "Resource forks", sure, but
POSIX Extended Attributes, of which ResForks are now simply a subset,
are a feature of several UNIX OSes now (and, unless you want to contrive
some entirely unique metadata storage scenario, necessary for the
implementation of ACLs). In Tiger, rsync simply knows about EAs and
Resource Forks come along for the ride. The bigger problem is that
there is no standardized POSIX (or UNIX03) API for EAs yet. FreeBSD
went one way, Linux went another way, and Apple (who needed some
additional parameter info) couldn't quite adopt either of the two APIs
as its own and went yet another way. Windows, of course, went yet
another way still with the NTFS Streams API (no surprise to anyone there).
It's true there is no complete standard, but there are similarities in
approach. It appears to be a sort of newtonian approximation process to
find the common middle ground. Apple appears to be trying the same
newtonian approximation process to a completely different function. The
treatment of the resource fork as an EA just doesn't fit with Apple's or
anyone else's EA model. The resource fork is like a stream that is
kinda sorta treated like an EA. TAdditionally, Apple has deviated in
the treatment of EA namespace and external representation of the EAs.
> So, in short, in the abstract sense there is general interest in having
tools like cp(1) and rsync(1) (to say nothing of the serializing
archivers like tar and cpio) deal gracefully with EAs, but it looks like
Apple's out on the end of the spear right now with respect to actually
deploying EA-based solutions. Linux has probably had EAs support the
longest (in the open source world, anyway), but it's not clear to me
that they actually use them for anything yet.
Redhat Enterprise Linux and Fedora have both had SELinux enabled by
default for quite a while now. SELinux uses EAs and ACLs rather
extensively and encountered and solved many problems along the way.
There are still many problems and Apple would do well to pay attention
to avoid the pitfalls they have encountered.
Rob
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden