Re: LEAST SURPRISE
Daniel Quinlan writes:
>> Sorry. What if you are mounting different partitions? You need to
>> mount with the BSD option each and every time you mount to have "least
>> surprise" or whatever.
Remy Card writes:
> And so ? I don't think that this is a real problem. [...]
Neither do I, really. I was mostly trying to point out that it isn't
possible to remove all the "surprises" from the system.
The "least surprise" argument is starting to get slightly over used,
especially since nobody seems to know what it really means or how to
quantify it.
>> Just change to another directory with the same permissions even
>> and then get different behavior [...]
> [...] If two partitions contain different types of filesystems, the
> semantics may be different (e.g. filenames lengths, access rights).
Good point.
>> I'd rather it was just the default Linux behavior, but it isn't going
>> to happen.
> Ah ? What makes you say that ?
Pessimism. :-)
> About two months ago, I have sent a query to the KERNEL channel of
> the Linux Activists mailing list. In this query, I had asked other
> hackers if the default behavior of the ext2fs should be BSD or SV
> like. There was no real answer so I did not change the default
> behavior of ext2fs but, if there is a need, I am willing to make
> this change in the next version (it is quite easy: I have only one
> line to change in the ext2fs kernel code). BTW, BSD semantics seem
> more logical to me so I think that I'll do the change someday
> anyway.
I agree, but having the semantics consistent across every filesystem
type is much more important to me.
>> I wish that we didn't have to even think about mucking around with
>> /etc/group. Anyway, the fact of the matter is that I'm not even
>> convinced that we want it. We will be getting good benefits, but
>> the drawbacks are undeniable. My mere question is: if the benefits
>> are so undeniable, why can't this change occur everywhere and not
>> just Debian?
> The change has not been made in other Un*ces and other Linux
> distributions because this change is simply not needed. Please see
> my mail in debian-user about the uid==gid proposal.
I did read it and it was a very well-spoken message. I have been
(mostly) wary of this change for some time now, but lacking of any
real arguments to dissuade its proponents (all two of them).
>> If the Linux community (or a significant majority) felt that
>> this is a good change and Linus agreed, could we have a low-level
>> change in directory semantics?
> Changing the directory semantics is quite easy in ext2fs (the code
> is already in and only needs to be flagged as the default behavior).
> Adding BSD semantics to other filesystems should also be obvious.
> I'd be able to change the directory semantics in every filesystem in
> less than one hour so if there is a need for it and no major
> objection, just tell me. BTW, don't forget that this is, IMO,
> unrelated to the common project scheme.
Maybe it is time to put it by the kernel channel again...
>> I don't think POSIX defines anything in this area, but I haven't checked.
> Posix allows both BSD and SV semantics, so either behavior is
> acceptable. FIPS-151 requires BSD semantics.
(Another reason to make the switch.)
Thanks for the worthy post, Remy.
Dan
--
Daniel Quinlan <quinlan@spectrum.cs.bucknell.edu>
Reply to: