Re: Case sensitivity in authz files

Mark Phippard <markphip_at_gmail.com> writes:

> That said, I did misunderstand the proposal. I thought it was
> suggesting we should just make things case-INsensitive. I thought
> that would be the wrong decision at this point.
>
> The file has always been case sensitive. So if you are just fixing a
> bug that was found, regardless whether it is contrived, I just do not
> see what the big deal is. There is no behavior change here. Fix it.
> Backport it to 1.7 or whatever. Why do we feel the need to announce
> anything or worry about the timing?

Making username handling case-sensitive seems like a good option, the
earlier behaviour was a very confusing mix of case-sensitive and
case-insensitive matching.

groups are a bit different. In the past they were simply
case-insensitive with the last group definition and access rule being
used. So

    [groups]
    foo = pm
    FOO = mp

    [/]
    @fOO = r
    @Foo = rw

defines a single group (= mp) and single access rule (= rw). Now that
is still a contrived example but something like this is more likely:

    [groups]
    Read_Users = philip

    [/]
    @Read_users = r

The new case-sensitive behaviour makes this file invalid while the old
behaviour granted read access. We probably want to avoid such a change
in a minor release.

-- 
Certified & Supported Apache Subversion Downloads:
http://www.wandisco.com/subversion/download