[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: (ITS#3639) Inconsistent access checking in back-shell?

> I forget why this check was added, but (as I recall) it
> was purposely added.  The reasons to why likely can be
> found in the commit log and/or -bugs/-devel list archives.
> I'll try to dig about later...

I've found back-shell/modify.c 1.23 -> 1.24 introduced this change;
there's no explanation, besides a comment that the same should be done for
back-perl and other scriptiong backends.  I think this is incorrect for
two reasons:

1) it is inconsistent with other backends and the difference is undocumented

2) the check is done using a dummy entry, built by placing the DN into an
empty Entry structure, which causes all access clauses that depend on the
contents of the object to behave unexpectedly; this is undocumented as

I've fixed this type of problems in back-sql recently, to restore a
consistent behavior; the same could be done in back-shell and so, but I
fear there might be issues related to fetching the required objects, and
it might not be worth the effort.  In any case, this point should be
clarified in the docs, regardless of how it's fixed.  I think we should
make access checking aware of the fact that it's working on a dummy
target; this should cause clauses that require the real data to take
appropriate measures (e.g. fail, or just do not apply, or at least issue a
warning that access checking might be incomplete or inappropriate for that

Maybe we need to discuss this in -devel.  In any case, there seems to be a
wide class of backends that do not honor access control as described in
the docs; there might be very good reasons for this, but they should be
made explicit.


Pierangelo Masarati

    SysNet - via Dossi,8 27100 Pavia Tel: +390382573859 Fax: +390382476497