Anybody working on removing sendmail from base?
cpressey at catseye.mine.nu
Wed Oct 1 12:42:39 PDT 2003
On Wed, 1 Oct 2003 11:38:00 -0700 (PDT)
Matthew Dillon <dillon at xxxxxxxxxxxxxxxxxxxx> wrote:
> :Are varsyms a part of DFVFS, or are they seperate and alongside it?
> :If they are seperate - is it a duplication of functionality?
> Separate and alongside it. Variant symlinks would be implemented
> in the core namei() and lookup() procedures, not in filesystems.
> Essentially they would simply be a reinterpretation of the
> contents of a symlink. e.g.
Ah. This jives with what I've gathered, doing a little Googling. FWIW:
I'm not sure that's what you were referring to, though.
Aside from the people who just don't like the idea of variant symlinks
(citing security or predictability or whatever), there seems to have
been a lot of opposition to them because "they're hard." When you go
changing namei() to support them, suddenly namei() has to be aware of
per-process information, and worse, it has to parse it, and apparently
people are a bit loathe make the parsing stuff in kernelspace any more
complicated than it already is.
(Of course, as I understand it, eventually namei() need not be in
kernelspace in DFBSD.... but that's probably beside the point.)
Interestingly - the only OS I could find mentioned that was known to
support variant symlinks was Apollo Domain/OS. *Are* there any others?
Maybe it *is* a hard problem...
I agree that DFBSD shouldn't need something as powerful as unionfs just
to support simultaneously installed different versions of packages.
On the other hand, if varsyms turn out to be just as much of a headache,
well, maybe a 100% solution is better than an 85% one?
More information about the Kernel