Re: Users & Groups specification proposal, draft 1

[gk4@us.ibm.com]

>> I'm not sure what the LSB process is, but this past January I posted a
>> there any formal acceptance criteria that I need to meet before I
>> begin SGML/SQL coding what I posted?

>The process is roughly that you just go ahead and check it in
>(especially if the text has already gone through the lsb-spec mailing
>list), and we worry about formalities later in the process.

>A few quick reactions (again, this is to
>http://www.debian.org/Lists-Archives/lsb-spec-0001/msg00056.html
>):

>* What's the purpose of standardizing /etc/passwd and /etc/group?  Is
>  the intent to provide a way for creating/modifying accounts?  Surely
>  we want something which is more general (e.g. shadow,
>  /etc/nsswitch.conf, LDAP, NIS, &c)?  We have useradd, groupadd,
>  &c....  Presumably I just misunderstand the intent here.

I was over documenting on the initial draft...  Only the APIs should be
documented, and not the actual internal formats of the /etc files.

Interestingly, I have never been on a system where I used the group
"password" field.

>* The text 'The "group" user database should only be read from
>  the following APIs' should also allow PAM (PAM might be implemented
>  by calling those API's, but we need to say that applications can
>  asume that PAM is doing an OK thing).  Likewise for passwords.

Noted.

>* What do SSM, URM, PRM, and RPM stand for?  Other standards, it
>  sounds like, but is there a more complete cite anywhere?

I was just being ganular on the BSD references (ie., System Manager's
Manual, User's Reference Manual, Programmer's Reference Manual).   I didn't
know if these will be eventually sorted by group according to their
end-user versus sys admin usages.

>* Why do you omit "newgrp" from the LSB?  There is a comment there
>  about chgrp but I don't see what newgrp and chgrp have to do with
>  each other.

You are correct.  I don't know why I have newgrp marked or commented that
way...  :-)

>* Why standardize rpc.rusersd, rusers, and rwho?

Just being complete on who was consuming uid/gid.  You are correct that
this should not be in the specification.

>In general this looks about right, though.

The "Users & Groups" section will be trimmed down alot once the "Commands &
Utilities" section is SQL/SGML coded.   The "Users & Groups" section does
not need to relist everything like I did...

Thanks,

George Kraft IV
gk4@us.ibm.com
512-838-2688; t/l 678-2688
IBM Linux Technology Center, Linux Standards