Can Any one give me some design documents of samba.

Jeremy Allison jeremy at
Fri Mar 30 09:24:19 GMT 2001

On Fri, Mar 30, 2001 at 12:52:45AM -0700, Matt Zinkevicius wrote:

> If someone out there uses it, then it must be supported. Of course that's
> completely orthoganol to whether we agree with it's existance :-)

No, that's not true. The only NT admins I've ever spoken to who actually
understood the difference between such a simple thing as "object inherit",
"container inherit", and "inherit only" were working for the US miliatary.

You don't *have* to support all customer demands, especially if they're
unreasonable (ie. NT ACL semantics) - you can just choose not to serve
the customers who insist upon this ("Let them eat Windows" :-). Does
your market research data tell you that all customers need NT ACL
semantics or is this just a case of a PHB saying "NT does it so we
must" ?  That way lies very poor code.

> I disagree. POSIX ACL's were designed for files. NT ACL's were designed for
> objects. It's only that we are looking at NT ACL's from a storage-centric
> viewpoint that we see them as overdesigned. Can you imagine POSIX ACL's
> being used to gate access on COM objects? Probably not :-)

I very much doubt that *anyone* has gated access to COM objects using
much more than users & groups with read and write. POSIX ACLs are perfectly
suitable for that of course. Anything more than that is unmanageable on
more than a trivial scale.

> No benchmarks. It shouldn't be too horrible since it doesn't hit my code on
> reads or writes. If you opening/moving/renaming thousands of files at a time
> then you'll probably notice a difference ;-)

I'm disagreeing with you, but I'd still like to see the patch :-).
Especially as you seem to have done most of the things I was
wondering about doing myself :-) :-).



Buying an operating system without source is like buying
a self-assembly Space Shuttle with no instructions.

More information about the samba-technical mailing list