[Gimp-developer] Re: plug-in distribution choices
[EMAIL PROTECTED] (2001-05-20 at 1535.17 +0200): Am I to understand that there is no recorded instance of this discussion? Well, let's start now, then, so that next time we can point to the mailing list archives. You can request IRC / mail logs, maybe someone has them. First of all a definition of the problem area: what are considered plug-ins? Everything that goes into the directory 'plug-ins'? Anything else? Some modules too, like the colour selectors, could be considered plug-ins. When talking about scripts earliers, I noticed that there is no clear way of distinguishing which scripts belong in the core distribution and which don't. I suggest we tackle this problem (first?). Scripts depend on the interpreter, so that would mean the interpreter should be in core app, which is not bad, but also not necessary. Scripts are plugins for plugins. My suggestion is that the following plug-ins belong to the core distribution: - those that perform a task that the GIMP should have provided for itself or will provide for in the future; I doubt Gimp will provide more, the idea is to have a small system where you plug things as needed, even GUI will be separate (and that for me is plugable). This general idea can be read in somewhere, this list archive probably. - those that will help other plug-in authors better understand how to write plug-ins; That should be in a devel package. Why do a user need a plugin that does nothing or near nothing? It is like the test script fu, lots of controls and you get a plain ball. Or like devel packages, users do not need to waste space with .h files that will never use. - those that will make the GIMP look good when compared to other raster image editors; and Then you include all those working, so you are sure you are giving the maximum you can. - those that perform a task the best in its field. I doubt there are repeated plugins, people do not code to have two, they patch instead. And if they did not know, they try to join the work, or deprecate the bad one. Can such a distinction be made? Yes, but you forgot: - plugins that are maintained. I think your point of view is what users want / need and the real thing is what volunteer coders can do. And from that comes the idea of reducing the number of core plugins, moving the rest to 1..N packages. That or somebody finds a way to keep all plugins updated (pay a coder with comunity money like with a Perl one? create a company to support it?). :] GSR ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
[Gimp-developer] Re: plug-in distribution choices
[EMAIL PROTECTED] (2001-05-20 at 1636.07 +0200): It turns out we need a plug-in manager. What functionality should such a beast have? Here's a list of things that came to my mind: [...] It should be able to download and install precompiled binary packages or download sources, compile and install them. A solution for the binary packages would be to use the binary packaging system provided by the distribution. Since there are a bunch of different distrbutions out there, this might be tricky. I would make a basic plugin handler so users can remove / add them to menus, if installed, and let the real task of getting the plugin to user / pkg system / whatever. If you use a pkg system, it can do it for you better than anything, if not, there is a make install target or a gimptool mode for that. I do not think becoming another Red Carpet is worth the time (which in place seems to be APT with GUI). Also it would mean secutiry audit for UID 0 routines. We all know how fun would be messages like I got new plugins, but my other account does not have them or it gets them, and then hangs while CPU and disk go 100% (compilation) or it seems hung, but after some minutes it says error, missing lib, and I have that lib (but not lib-devel). If people with knowledge can have problems, imagine the rest. So I would split the system more at source level, so you get x groups and build install them (or make distro pkgs then install) following an order, like you can do with GNOME, ie. ATM I already use that (with x=2): gimp and gimp-data-extras. GSR PS: OTOH it would be nice if Gimp could read post mail news. And IRC too. /me ducks runs. ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] Re: plug-in distribution choices
Hi, Guillermo S. Romero / Familia Romero [EMAIL PROTECTED] writes: I would make a basic plugin handler so users can remove / add them to menus, if installed, and let the real task of getting the plugin to user / pkg system / whatever. If you use a pkg system, it can do it for you better than anything, if not, there is a make install target or a gimptool mode for that. I do not think becoming another Red Carpet is worth the time (which in place seems to be APT with GUI). [...] So I would split the system more at source level, so you get x groups and build install them (or make distro pkgs then install) following an order, like you can do with GNOME, ie. ATM I already use that (with x=2): gimp and gimp-data-extras. actually this is my opinion too. I'm not convinced that we should try to deal with binary packages. This is one of the ideas that has come up and that I remembered, but I second your thoughts about it. IMHO the best solution will be to have a bunch of standalone source packages that follow some well-defined rules. One rule should be that there needs to be a file that describes the package and all its components that can be used by the plug-in manager and by the next generation plug-in registry. Ingo had an XML format in mind and I was hoping he would post a proposal to this list, but I haven't seen it yet. For the moment we want to keep all plug-ins in the core package until we have ported Gimp to Glib/GTK+-2.0. This is because we think that a lot plug-ins will be ported very easily and having them all in one place might ease this task. Once the port is done (and we are going to start it very soon now), we should think about moving most of the plug-ins out of the core CVS module. Hopefully we have made up our mind on the new plug-in system until then. Salut, Sven ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] Re: plug-in distribution choices
Sven Neumann wrote: Hi, Guillermo S. Romero / Familia Romero [EMAIL PROTECTED] writes: So I would split the system more at source level, so you get x groups and build install them (or make distro pkgs then install) following an order, like you can do with GNOME, ie. ATM I already use that (with x=2): gimp and gimp-data-extras. actually this is my opinion too. I'm not convinced that we should try to deal with binary packages. snipped... If by deal with binary packages, mean make binary packages, I don't think we have the resources. Of course, we can't be indifferent to package makers of the distribution companies at the principle and practices level, since the workings of the plug-in manager will have an influence on how to make packages. IMHO the best solution will be to have a bunch of standalone source packages that follow some well-defined rules. One rule should be that there needs to be a file that describes the package and all its components that can be used by the plug-in manager and by the next generation plug-in registry. I also think this general approach is correct, but the design of a package manager would take a great deal of care (albeit, very useful). This particular rule readily turns into a protocol encoding a plug-ins' dependencies on other resources, (other plug-ins, modules, libraries, interpreters), its preferences in a menuing system, version level requirements, etc. For the moment we want to keep all plug-ins in the core package until we have ported Gimp to Glib/GTK+-2.0. This is because we think that a lot plug-ins will be ported very easily and having them all in one place might ease this task. This would also be an ideal pipeline to inventory existing plug-ins with an eye toward how well they can be adapted to a package manager, what constraints they would place on the design of a package manager, establishing weak/strong dependencies among plug-ins, and deciding about functional groupings (i.e., logical packaging). Once the port is done (and we are going to start it very soon now), we should think about moving most of the plug-ins out of the core CVS module. Hopefully we have made up our mind on the new plug-in system until then. I think the Gimp could also use a Plugin Maintainer on par with the Core Maintainer to midwife this effort. Be good, be well Garry ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
