Path: utzoo!attcan!uunet!husc6!bloom-beacon!ATHENA.MIT.EDU!swick
From: swick@ATHENA.MIT.EDU (Ralph R. Swick)
Newsgroups: comp.windows.x
Subject: Re: Resource management on 11.2
Message-ID: <8809161430.AA02552@LYRE.MIT.EDU>
Date: 16 Sep 88 14:30:11 GMT
References: <8809160015.AA14382@decwrl.dec.com>
Sender: daemon@bloom-beacon.MIT.EDU
Organization: DEC/MIT Project Athena
Lines: 58


     Date:  Thu, 15 Sep 88 17:09:12 PDT
     From:  sequent!gssc!davec@decwrl.dec.com

        Why is there a separate .Xdefaults and .Xdefaults-xxxx file, and why
        are they treated differently?

        Why does the use of xrdb exclude the use of .Xdefaults?

Here's the general philosophy:  user preferences are most usefully
associated with _screens_, not with hosts (or filesystems, or...).
Clients connecting to a particular Display from anywhere in the
network should get the user's preferences.  This can only happen
when they are stored in the server, thus the implementation of xrdb.
The fallback to .Xdefaults is only out of habit 'cause there have
been lots of users of X prior to the existence of xrdb and we didn't
want them to loose from day 1.

     2) .Xdefaults-xxxx quickly reaches the 14 char filename limit of sysV
        and other OS's.

Which is exactly why it's an environment variable.

        Are any plans in place to fix this in R3 or later releases?

nope.

     3) I don't understand the purpose of .Xdefaults-xxxx files.

The XENVIRONMENT option is a last-chance way for users to specify
customizations on a per-process basis.  Providing a fall-back filename
may be a bug; it adds confusion, rather than clarity.  The assumption
was that a frequent use of per-process customizations was to
modify the user's server defaults slightly (e.g. different backgrounds)
according to the host from which the client (e.g. xterm) was connecting.

     4) /usr/lib/X11/app-defaults is a directory of files where the file name
        is a classname.  For xterm, the classname is XTerm.  Each client has
        it's own classname (as far as I can tell.)

Each client defines the class of which it is a member.  The class may
include more than one application.

					  This is not documented
        well.

guilty as charged.

           For instance, given the set of clients from MIT, what are the
        given classnames?

they all use the rule of "capitalize the first letter, or the first and
second letter if the first is 'x'".

			  Is there a way for the user to specify classnames
        at runtime?

No.  The class name is reserved to the application developer.