On 5/23/08, Ciaran McCreesh <[EMAIL PROTECTED]> wrote:
>  > So how do you manage large network with paludis?
>  > You compile from sources at every endpoint?
>  > Say you have 1000 workstations with different applications installed
>  > at large groups, do you expect network admin to compile each package
>  > over and over on all workstations?
>
>
> No, I expect them to use system images. And just install every
>  application everywhere, since disk space is cheap.

It is great that package manager forces a specific solution.
But unfortunately there are some different views and I don't
understand why you force a specific solution (network implementation).

Why do you decide for me if I want to pay extra for disk? And if I had
laptops, that cannot enlarge disks easily?
How do you know (dirk) that the compunter is always connected to the network?
And if I don't want some application available for some of the workstations?
Why making network administration harder than it should?

This is not for you to decide for users.
You can however tell me that you are not going to provide binary
packages because you think it is not important to you guys, but don't
reason this as it is better for users.

Looking at large scale installation is somewhat you did not do. The
quickpkg provides good solution for this, you can learn from its
motivation, as it provide the laking link that is needed for large
distribution of packager output.

>  > >  > no phases for package developers (ebuild)
>  > > Again, at some point, unless we decide that implementing them will
>  > >  stop developers from fixing their bad development habits. But
>  > > again, they're not an essential feature nor even particularly
>  > > useful.
>  >
>  > It is great you want to force quality, but I don't understand why you
>  > keep mentioning that this stop developers from doing good job... As I
>  > wrote you in ticket#580, you cannot do good job if you don't verify
>  > the outcome of every phase. This does not mean you don't check the
>  > outcome of the whole process.
>
>
> Uh, the correct thing to verify is the process as a whole.

The right answer can be given even if the logic of ones though is wrong.
If we go into details, there are white box checks and black box
checks, you are forcing developers to use black box only, while the
white box tests are the ones that finds most of bugs.

Again, you force your QA view on users, although it may be incorrect,
without providing the ability of users to use the process the choose.
Had you said that you don't provide phases as it complex for you to do
this, I would have understood, but if you do this just to force your
(wrongly) view of QA management, I think this is incorrect.

> Why would it make migration any easier? You might as well install
>  dselect for all the difference it would make...

Package management is the major components, as you can see from our discussion.

Thanks,
Alon.
_______________________________________________
paludis-user mailing list
[email protected]
http://lists.pioto.org/mailman/listinfo/paludis-user

Reply via email to