[Freeipa-users] UID 999, not possible?

[Nathan Kinder]

On 07/03/2012 12:38 AM, Petr Viktorin wrote:
> On 07/03/2012 05:55 AM, Nathan Kinder wrote:
>> On 06/29/2012 07:10 AM, Petr Viktorin wrote:
>>> On 06/29/2012 03:55 PM, Alexander Bokovoy wrote:
>>>> On Fri, 29 Jun 2012, Petr Viktorin wrote:
>>>>> On 06/29/2012 03:04 PM, Alexander Bokovoy wrote:
>>>>>> On Thu, 28 Jun 2012, sysadmin at noboost.org wrote:
>>>>>>> Hi All,
>>>>>>>
>>>>>>> Is there a weird restriction to UID 999 in ipa, as IPA keeps 
>>>>>>> changing
>>>>>>> the UID when I add a user with that number? (I've already 
>>>>>>> checked the
>>>>>>> UID isn't in use)
>>>>>> We use 999 as a marker for DNA plugin. UID/GID 999 is replaced by
>>>>>> an allocated one with the help of the 389-ds plugin
>>>>>> http://directory.fedoraproject.org/wiki/DNA_Plugin
>>>>>> http://docs.redhat.com/docs/en-US/Red_Hat_Directory_Server/9.0/html/Deployment_Guide/Defining_Dynamic_Atrribute_Values.html#about-dunamically-assigning-attribute-values 
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>> The documentation mentions that the magic value can be a word
>>>>> ("magic"), or it doesn't have to exist at all (it's added for
>>>>> objectClass:posixAccount entries). Is there a reason IPA is using 999
>>>>> here?
>>>> uidNumber and gidNumber field use integer value syntax:
>>>> OID value: 1.3.6.1.4.1.1466.115.121.1.27
>>>>
>>>> OID description:
>>>> Values in this syntax are encoded as the decimal representation of 
>>>> their
>>>> values, with each decimal digit represented by the its character
>>>> equivalent. So the number 1321 is represented by the character string
>>>> "1321".
>>>> So, you can't have string there that does not evaluate to integer.
>>>
>>> That's true, but according to the documentation you linked,
>>> uidNumber/gidNumber syntax doesn't matter.
>>> The dnaMagicRegen field is in fact a DirectoryString. I assume the DNA
>>> plugin sees and modifies the value before it's validated as an integer.
>> I wouldn't trust this, as DNA was initially designed/implemented before
>> we added syntax validation to 389.  DNA was also written to be able to
>> work with non integer attributes, where values have some sort of prefix
>> followed by an integer (such as "user1", "user2", etc.).  For this
>> reason, dnaMagicRegen was left as "Directory String" syntax.  I
>> personally feel that it is safer to have the magic value be
>> syntactically valid for the attribute that DNA is configured to 
>> generate.
>
> Best go with a negative number then.
> The DS docs should be updated if you don't trust what they say, though.
I should have been a bit more clear.  I believe that the core 389 DS 
code does handle replacing the magic value before the syntax is checked, 
but we have encountered issues with client applications trying to 
enforce syntax before the server receives the operation. An example of 
this is the 389-console application (which FreeIPA doesn't use).  The 
Console knows that the uidNumber attribute is defined to use the Integer 
syntax, so the UI field is validated before it can be submitted to the 
server.  Other client applications may do similar things by looking up 
the schema definitions on the server and trying to do some client-side 
validation.  This sort of behavior prevents the ability to trigger DNA 
if the magic value does not meet the syntax requirements of the 
attribute.  Again, this may not affect FreeIPA, but I wanted to provide 
a bit of background.
>
>
> On 06/29/2012 04:23 PM, Alexander Bokovoy wrote:
> > Looks like you are right:
> > 
> http://comments.gmane.org/gmane.linux.redhat.fedora.directory.user/10641
> >
> > We would have issue on our side when using non-integer value as Int()
> > parameter does not support non-integer values. However, we could select
> > some negative value as default one and use the same value for DNA
> > configuration.
>
> The value can be optional, the server can fill in the default if it's 
> not received from the client.
>