Re: (ITS#3613) globbing support for slapd config's include statement

[hyc@symas.com]

Kurt@OpenLDAP.org wrote:

>I note that this patch, as submitted, is somewhat dangerous.
>There are existing include directories whose first argument
>is not expected to be considered a glob, and which if
>were treated as a glob, would lead unexpected files being
>included.  For instance,
>        include *.schema
>
>might have been intended to match the file named '*.schema',
>not all the files whose names end with '.schema'.
>
>It would be better to introduce include globbing in some
>backwards compatible way (e.g., another directive).
>
>Also, I see that code does include any safeguards (e.g.,
>checking that the path refers to a file, as opposed to a
>directory).  Nor does the patch include any documentation
>updates.  Nor is the patch conditional on the availability
>of glob(3).
>
>And, as Ando noted, HEAD is much different here than RE22.
>
>That is, this patch needs work.
>
I am inclined to reject this patch regardless of any reworking. One of 
the key points a lot of people seem to miss is that schema files have a 
specific dependency order; references must always be defined before they 
are used. A feature that allows one to say "just load all the schema 
from location X" would tend to further obscure this crucial point. On 
the other hand, with back-config in 2.3, the behavior will already be, 
essentially, "load all the schema from 
LDAP_SYSCONFDIR/slapd.d/cn=config/cn=schema" but as we will have 
well-documented rules for the structure of the cn=config hierarchy, its 
behavior will be precisely defined.

Since the provided patch is against 2.2 and 2.2 is feature-frozen, it 
cannot be accepted there. Since 2.3 largely supersedes it, I don't see 
any reason for an adapted patch to go into HEAD.

-- 
  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc
  Symas: Premier OpenSource Development and Support