Re: [G2R] Why is bonobo-config a 2.0 core library?

[Michael Meeks <michael trna ximian com>]

Hi Stephen,

On 23 Nov 2001, Stephen Browne wrote:
> this is why I was hesitant to send that mail in the first place. I
> guess I knew someone could not resist the temptation to turn this into
> a b-c vs gconf merit war.

        :-) I didn't discuss any of the merits of bonobo-conf in any depth
- but I just don't see you argument that "no one uses it" as being
particularly valid, whereas "it's useful for ..." is more valid; library
APIs are created to be used by applications, and since we're not shipping
many applications yet it seems reasonable to me that no one is using it.

        eg. I very much doubt that anyone is using libgnomeprint yet. Does
that mean we should drop it from the platform ? I assume not.

> I'm looking at facts here, not architecture, not personal opinions of
> other peoples ability to design or code and I'm not looking to be
> educated in the whole area of configuration.

        Which is why I'm not trying to educate you - and resurrect a long,
and grossly ill informed debate - the conclusion of which is that Havoc
and I agree that there should be a choice of APIs to access configuration
but a single backend. It seems however that the mis-information from the
original debate is all that is remembered and is still slopping around in 
people's heads.

> One less library delivered means one less public API I have to
> document as part of the platform and one less API I have to
> support=>cuts costs.

        Well - the API is substantialy the bonobo property bag API - which
we already ship - plus the bonobo moniker API - which we already ship -
plus the knowledge that the conf: moniker points into the GConf database;
that's hardly a huge ammount of new API to document.

> there is no need to be nasty about it.

        And indeed - I don't want to be nasty about it - I'm sorry you
took it that way; partly I took it badly because I need to do some serious
pruning inside bonobo-conf myself and make it more suitable for what we
need; so partly you are poking your finger into a large community sore,
but also into a personal one.

> > I won't be there; sadly since I'm on holiday - so go for the
> > discussion - lets make it really uneducated, like the last spate !
>
> This is just nasty!  Im sure the release team didnt enjoy reading that
> comment.  I do hope you enjoy your holiday though.  My wife is at me
> to book a weekend away somewhere so I must get that done soon.

        Well - the thing is your arguments initialy sound convincing, but
are ( I feel ) based on misunderstandings:

                * No one uses it,
                * The API will need documenting

        But in fact they are not such good arguments - as discussed above,
and some apply to libgnomeprint[ui] and other APIs in the platform. Thus
for a discussion to be based on them would be ( I feel ) really uneducated
- and I was in a hurry to get away and could not address this before I
left, nor respond to the issue if/when it was discussed at the release
team - which is frustrating.

> GNOME 2.0 which is due to beta in the first week on Dec and the
> agreement is 'use gconf'.  Even Nat agrees :)
>
> http://mail.gnome.org/archives/foundation-announce/2001-November/msg00022.html

        Sigh; if you mean:

    " ...we need one configuration mechanism (and it should be GConf);"   

        Then I agree with Nat's statement; and bonobo-conf allows this -
simply through a different API; again in agreement with what we had
decided collectively.

        In conclusion:
        
                My feeling is we should ship bonobo-config as agreed   
                My feeling is bonobo-config needs work ( as does much of 
                Gnome 2.0 ) and I will get to it soon.

        Regards,

                Michael.

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot