Re: Hurd and Unix/Linux and Plan9 features

[olafBuddenhagen]

Hi,

On Tue, Feb 06, 2007 at 05:00:53PM +0100, Alfred M. Szmidt wrote:

>    I think it would be useful to have a link type in the Hurd that can
>    fully emulate hard links in the above example (by transparently
>    switching the firmlink to a hard link if the primary link goes
>    away), but firmlink doesn't do that -- and sadly, I see some
>    fundamental problems, with probably make such a behaviour
>    unfeasible :-(
> 
> I see such problems as well. Do hard links even work on GNU?

Of course they do. Why wouldn't they?

Problem is that they are of limited use. Hadlinks are good enough as
long as only traditional disk files are involved -- but leaving this
"safe" ground can happen very quickly, e.g. when using unionfs. Using
hardlinks (or something that behaves like hardlinks) seems much more
interesting if you can link stuff between different translators, so you
could for example move pseudo-files between translators without
explicitely copying the contents... A link-unlink based move operation
allows for some interesting semantics, which seem hard to express
otherwise.

I have some ideas for a mechanism that would allow to do such
inter-translator linking, but it seems rather complicated and there are
some possible failure modes that I'm not sure how to handle sanely. All
in all, I'm not sure it's really worth the trouble. Probably I will only
find out once I start doing fancy stuff with translators...

> One could, I suppose, use a similar hack to how symlinks are handled,
> I think.

No idea what you mean :-(

-antrik-