Kernel package and modules

[Rod Whitby]

Andy Green wrote:
> Somebody in the thread at some point said:
> 
> | |> I would like to propose we radically simplify how we deal with
> | |> modules, simply by pushing all the ones we create into the main
> | |> kernel package. This has one disadvantage, size on storage, but
> 
> | |> What do the folks who handle this packaging think about this proposal?
> | |>
> | | Or we could just RRECOMEND kernel-modules which will drag in all
> | | modules. See we are ahead.
> |
> | What is "kernel-modules" then, a metapackage for all modules?
> |
> | I can't "see we are ahead" if nobody tells me what is happening.
> 
> Assuming this is so, why do we generate a bazillion pointless module
> packages still if having the kernel package brings 'em all in anyway,
> and they will ALL be regenerated at each kernel package uplevel?
> 
> How is that "ahead" of just making a single kernel package that has the
> modules in too?

The reason why that is done in OE (as opposed to why OM should or should 
not do it) is because OE is used for devices with 8MB total of internal 
flash for kernel, modules and full rootfs.  In that environment, you 
want them to be as granular as possible so you don't waste jffs2 space.

Of course, the GTA02 is not such a device, although some people may want 
to use jffs2 space for applications rather than a bazillion kernel 
modules that they will never use ...

-- Rod