I'll reply inline...

Garrett D'Amore wrote:
> On a separate note, its unclear to me whether SUNWemacs-nox is a sound 
> approach for Solaris.
> 
> It presents potential difficulties, as there is no precedent (that I'm 
> aware of) for having the same binary program (/usr/bin/emacs) 
> distributed by different packages.  It potentially raises challenges for 
> patching, and other sustaining efforts, and I'm not entirely certain 
> that this has been thoroughly considered yet.

That's not the case. /usr/bin/emacs is delivered by exactly one
package (SUNWemacs). It is a shell script, that runs the best
actual emacs binary available. The user has a choice of emacs binaries,
and is free not to install the ones that are not of value.

The script is in the case materials.

Those binaries come from multiple packages, but each delivers under
a unique name (/usr/bin/emacs-x, /usr/bin/emacs-nox, and possibly
/usr/bin/emacs-gtk).

I don't think this breaks new ground or creates a new problem.


> Notably, SUNWemacs-nox would be a strict subset of functionality, and 
> its main (and perhaps only) benefit would be to reduce the size of the 
> installation image (e.g. for minimization environments) so that it is 
> not dependent on X11.

It isn't really much smaller than an X11 version either, thanks to
sharable libraries being used for X11. I don't think I'd use it
either, but there are reasons to include it:

        1) Optional package: Don't want it? Don't install it.
        2) Linux does deliver it, using this same approach
        3) IPS opens doors to minimization that do not exist today,
           and one obvious such minimization is to leave off X11.

I make no claim that this is particularly valuable today, but see it
as a placeholder for the future, and one that doesn't those who
don't want it. I am trying to establish the way this package will
be delivered in the future, as well as today.


> 
> Also notably, emacs runs fine in a tty without X11 running, even when 
> compiled with X11 support.

I completely agree, except for the case where you have no X11 libraries
installed. An emacs compiled with X11 support will no run without those
libraries, even if they are not installed.

As I said above, we're following established patterns with this.


> Since there is not yet any standard for minimizing Solaris -- at least 
> as distributed by Sun (and indeed Solaris is already rather large), can 
> I humbly suggest to the project team that they might want to consider 
> dropping this portion of the project.  Other distribution builders that 
> want to deliver a system without X11 will have a larger task ahead of 
> them anyway, and can easily repackage emacs to their own taste if so 
> desired.

Since the package is optional, it would be reasonable to simply
not install it.

> 
> As an alternative, perhaps delivery of an emacs called 
> "/usr/bin/emacs-no-x11" or somesuch might be a reasonable way to avoid 
> the confusion from delivering different object files in the same 
> location, although I still personally feel that there is little merit to 
> the notion of a version of emacs without X11 support.

Covered above --- they're not delivered to the same name. The binary
is actually named 'emacs-nox'.


> (As an aside, in Linux distributions, this is used for environments such 
> as recovery floppies, and in ridiculously tiny embedded environments.  
> Solaris is not targetted at any of those environments -- I'm talking 
> environments with < 16MB of storage here!  In Solaris, X11 is a core 
> depenency, IIRC, and I'm not sure there is an easy way to install it 
> without it.)
> 
>    -- Garrett
> 


I make no claims about the future of minimized Solaris, but still feel that
this decomposition is reasonable. It puts no additional barriers up to
minimization, and it follows the way other systems work.

Personally, I will probably install SUNWemacs, and SUNWemacs-x, but
not the other two.

Thanks...

- Ali


Reply via email to