Re: (ITS#3890) Back-sql add ignored fail_if_no_mapping option

[ando@sys-net.it]

cjones@simdesk.com wrote:

>Full_Name: Chad Jones
>Version: 2.3.4
>OS: Redhat 9
>URL: ftp://ftp.openldap.org/incoming/chad-jones-050725-6.patch
>Submission from: (NULL) (65.65.189.5)
>
>
>Back-sql's add functionality seemed to return an error even if the
>fail_if_no_mapping option was setup for that backend.
>  
>
I think this patch should be rejected, because the fail_if_no_mapping 
feature means that writes should/should not fail if writes cannot be 
mapped; as soon as a write (i.e. add) procedure for a certain type is 
provided, if anything else required for writing (e.g. the capability to 
collect values that must be deleted for a "replace") must be present.  
The capability provided by the proposed patch would implement a sort of 
"ignore errors" which is not the intended behavior.  If anything like 
that may be of use (personally, I doubt; fail_if_no_mapping should be 
the default, and there should be an option "dont_fail_if_no_mapping") it 
should be suggested as a separate feature, with its own configure flag, 
or as a set of failure levels: "fail {never|no-mapping|always}".  The 
requirement of "fail_if_no_mapping", with "don't fail" as the default, 
is given by the need to allow writes when attributes like operational 
and accessory stuff which is not necessary for operations cannot be 
mapped to the RDBMS, or simply is deemed not essential; it results in a 
break in the LDAP data model, so it should be used with care.

p.


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