[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: remove fedora-usermgmt?



On Fri, Mar 09, 2007 at 05:26:02PM +0100, Michael Schwendt wrote:
> On Fri, 9 Mar 2007 16:19:47 +0100, Axel Thimm wrote:
> > On Fri, Mar 09, 2007 at 01:31:13PM +0100, Michael Schwendt wrote:
> > > Predictable means you can keep the uid/gid constant,
> > 
> > in a floating window.
> 
> Eh? What do you mean with "floating"? That you refuse to keep the base
> value constant? Well, there are lots of other ways how to damage
> installations.

Ehem, the floating window *is* the core feature of this package, but I
agree that it's a bug called a feature ;)

> > "Constant" is the definition of a fixed uid. If there is need for a
> > fixed uid, ask for one (yes, there _seems_ to be currently no space, but
> > that is another issue), if not use useradd -r.
> 
> And with useradd -r, how do you get the same uid/gid for a package on all
> installations, when you want that?

Technically by just adding the uid/gid as an argument to
useradd/groupadd.

From a distibution POV you need to register a fixed uid/gid first, of
course. And the entity that is in charge of handing out the fixed
uid/gid will consider your application and accept or reject it.

> > > but still have an influence on where that is within your range of
> > > values. Everytime you install a package again on a machine under
> > > control of a configured fedora-usermgmt, the package allocates the
> > > same uid/gid.
> > 
> > sure - oops, the admin forgot to configure fedora-usermgmt on machine
> > number 23. Now all uid/gid are messed up.
> 
> In the same way you can install the wrong distribution on machine number
> 23 e.g. because you insert the wrong media. ;) fedora-usermgmt setup can
> be made available at installation time.

No, forgetting an additional imposed configure step is different than
inserting your Vista CD.

In fact if *all* package were to use this scheme, you would have no
chance in configuring this before the first package makes it on
disk. Or do you want anaconda to ask "please insert a random uid/gid
base, don't worry if you don#t know what this is all about, just pick
your lucky number" ;)

Yet another issue with this scheme, how cool, one only needs to think
it through to find one flaw after the other.

> > From any POV I look at it, this design is flawed ...
> 
> Hyperbole.

Blind man talking?

I wish we had a voting in the EPEL sig on that. Sigh.

> > Or at least present a counter-example, where a package needs it.
> 
> You can't think of any environment, where predictable uids/gids make
> sense?

Gimme the example. No more theories and rhetorics, please. :)

> > And then please explain how it can need a fixed uid/gid and still
> > have survived that long in the
> > fedora-usermgmt-defaults-to-useradd-r setup.
> 
> Because default behaviour is transparent and just like "useradd foo".

Which means that the package didn't require anything more to begin
with. QED.
-- 
Axel.Thimm at ATrpms.net

Attachment: pgpJWQnAVwnLV.pgp
Description: PGP signature


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]