[plt-scheme] Unix install

Eli Barzilay wrote:
> On May  2, mvanier wrote:
>> I find this discussion somewhat surprising.  I run PLT scheme on
>> Linux, install into /usr/local/plt, and all I have to do is to set a
>> few shell variables, specifically PATH, MANPATH, LD_LIBRARY_PATH,
>> and PLTHOME.
> 
> The unix counter argument would be that it is unrealistic to have
> *PATH variables that include a huge number of application directories.
> 

Yes, this is a problem.  My understanding is that shell variables can be up to 
65536 characters long on most shells.  If this is the case, it gives you a lot 
of leeway but not a sense of security.  Frankly, there's no good reason for this 
that I can see.

> 
>> Not exactly a huge burden IMO, especially since now that they're set
>> in my shell initialization script, I can install new versions with
>> no extra effort.  I've never had any problems with this.  You could
>> argue that it's worse if you have to install for a large group of
>> users, but not much.  I frankly think the traditional Unix
>> installation conventions are a relic that should be overhauled (and
>> I'm not some kind of Windows or Mac advocate either; I'm a hard-core
>> Linux user).  On my machine, /usr/local/plt has 6832 non-directory
>> files in it.  Personally, I like the feeling of knowing that I can
>> find everything PLT-related under one directory.
> 
> I'm in this camp too.  A better solution (IMO) would be a single
> APPPATH that specifies a directory of applications, where each one has
> subdirectories for man/lib/include etc.  But that's unrealistic.
> [When I searched for ways to get to the current executable filename on
> Unix (no portable way to do that), I consistently got to pages where
> "Unix Gurus" will go on lengthy discussions explaining that it's
> usually due to people who want to keep extra files in the same
> directory (for single-directory distributions), why you should conform
> to the standard way instead, and why you're an idiot for trying to
> fight it (that last argument often degrades to "decades of unix can't
> be wrong").]
> 

Yes, an APPPATH would be vastly preferable.  People who advocate the Unix 
conventions should spend some time reading about the Plan 9 operating system, 
and realizing that the original authors of Unix realized that it had many flaws 
which needed to be fixed (though I don't think they addressed this particular flaw).

Mike