gstep-make 2.0 Re: gstep-base Re: [OGo-Developer] OGo InvoiceApplication

[Sebastian Reitenbach]

developer@opengroupware.org wrote: 
> On 02.01.2008, at 16:00, Sebastian Reitenbach wrote:
> > This is exactly what my opinion is about ogo packages.
> 
> 
> Oh common, you just said that you always build from source anyways! ;-)
Yeah, usually I do, at least for ogo, because I want to develop, and need 
the sources anyway.

> 
> The far majority of OGo users *is* using packages, and often is not  
> experienced in compiling software.
Yes, therefore a GOOD documentation of that process, and all that nifty 
stuff  that the meta-packages provide, could be done via a make call, as 
e.g. that request tracker is doing it. then there is no point. And I know, 
if there are no rpm's for a software, then there are people that will not 
try the software at all, despite how easy it is to install it from source.
That is the reason why I want to create ports for OpenBSD, they are not for 
me, I want to get people to use OGo on OpenBSD. Afaik, I am the only one 
using OpenGroupware on OpenBSD. The reason is just that there is A) no port 
of it, and B) it will not copmile and install, without the bunch of patches 
I have to apply to make that happen.

> There are platforms which have decent packaging systems (eg Debian/ 
> Ubuntu) and its stupid not to use them for initial installation and  
> keeping systems up to date.
> Anyways, this obviously is additional work ...
> 
> > I just have the fear that when we maintain own stuff, I fear that we  
> > get
> > stuck with one version, and then maybe later in the future, when we  
> > want to
> > upgrade a whole bunch of versions, then facing same problems as  
> > right now.
> 
> You are starting to understand why using gstep-base as the primary  
> basis is not as cool as it sounds. Its OK if you maintain one or a few  
> systems, but its havoc if you provide shrinkwrapped software to  
> endusers (with hundreds or thousands separate server installs!).
> As a user I cannot install business software which provides no  
> appropriate upgrade management!
> 
> And again: if you live at the bleeding edge this automatically implies  
> that you cannot use system packages.
Well, I maintain my own ports for gnustep-unstable, for boxes where I need 
it. In the official system ports, there are always the newest gnustep stable 
versions, short time after they get released. most of the time I update 
them, and send them to the maintainer that he updates them into cvs.

> 
> > Trying to keep up with the new versions when they are released,  
> > would make
> > updates easier as the differences will not be that much, if at all.
> 
> 
> The differences have been *significant* between releases in the past,  
> I don't see why that would change in the future. Eg even minor changes  
> in the KVC code can affect OGo and SOPE badly.
> Check the archives of the gnustep-discuss list, I think I once sent a  
> summary to outline that point (highlighting major incompats between  
> minor versions).
I think I can remember that thread.
> 
> BTW: I think the GNUstep guys pretty much agree with me on this :-) In  
> fact *we* could probably contribute a stable GNUstep release branch.  
You mean a second gnustep-base stable? there exists already a gnustep-base 
stable version? right now it is 1.14.2 I think, the unstable stuff is at 
1.15.2.

> It would be our task to manage the stable GNUstep branch (add ABI  
> compatible fixes, etc).
> Anyways, as posted before _I_ won't invest that time anymore for  
> various reasons (ObjC 2.0 being one of them).
And then instead, keeping the existing ogo stuff compiling against the 
(dead) libFoundation? Then this will make the existing objective-c ogo die 
sooner than later I think. At least, when ogo will be compatible to 
gnustep-base, it might attract one or another people to develop, as I have 
seen happen to sogo. But as long as ogo depends on libFoundation, this will 
very unlikely happen.

> 
> > The Novell distris would be the only Linux distributions, where I am
> > interested in having packages too.
> 
> Si. But this is not just about us. (eg personally I only care about  
> Debian, and maybe Ubuntu ...).
ah, that is good, so there would be one caring about good packages for these 
distributions ;)

cheers
Sebastian

> 
> Helge
> 
> PS: all this is not intended as a blaim on GNUstep! (I think its easy  
> to get my comments wrong ;-) Technically its great stuff and  
> especially gs-make and gs-base are in very good shape. But the GS  
> developers (I know) have widely different requirements wrt deployment,  
> and in some respects they use the software in a very different way (eg  
> no KVC is the classic).