Preamble:

TINY's target is  relatively small boxes but also people who are not
computer scientists ad whose main motivation is not learning but doing
some real work.  This motivates for the install being relatively
inflexible: the lesser questions the installations asks, the lesser
the user will be confused.  The way we envision future extensions
would also be rigid but plug and play: ie if you want to install the
Office suite, you just insert media and when finished you will find
it works.

Here are some improvements that could be added:

1) A better configurator for X and an integrated and fool proof X
installation: it should not be possible for the user to not install
the X server he needs.  Here we could pick an existing X configuration
tool but the install for X will need a major face lift, it would
probably be easier to completely redesign it.

2) TINY needs a printer configurator.  This is very important: remember
TINY is for real work!

3) Three issues need to be investigated: installation from a Zip, from
external CDROM readers and through a LAPLINK cable.  Installation from
a Zip needs reviewing what happens when all the installation is on
single media instead of many small ones (like for a floppy
installation) otherwise there is no real problem and after that it
will be straightforward to allow installing from similar media types.
Installing from CDROM will need to take into account that many
computers in TINY's niche had old pre-IDE CD readers and in addition
you had to deal with troublesome IRQs or memory ranges.  IMHO we would
probably be forced to move to a modular kernel (there are other
reasons for doing so) but again this would require modifying the
installation.

4) Since TINY aims to be a distribution for real work we need to build
an office suite made of light-weight software.  Last time we checked it Ted
looked like it was we needed as Wysiwyg word processor but it was
still rather buggy.  Need to be rechecked.  We could also investigate the
use of XFCE as a more palatable alternative to mere use of Window
manager.  However here we have to consider how much additional memory
that would need.

5) An old version of Netscape will fill the role of navigator.
Unfortunately these old versions use a.out format and this means
additional memory consumption.  Since at that time Netscape did not
read mail and news that means we would have to investigate what mail
and news readers are available.  We also need a program for easy
configuration of PPP access (eznet and its xeznet pal seem good
candidates).   These components could  form an Internet suite.

6) A way to save memory would be to rebuild X and disable support for
features having little pratical use(*).  Another way to save memory would
be to rebuild the SVGA server and disable support for cards like the
TNT2 who are obviously out of our niche.  This would be needed and
useful only as long as TINY stays with XFree 3.
(*) Example: multihead support

7) Moving to SysV init sequence.  The BSD init sequence used in
Slackware is easier to understand by a human but it makes difficult
both the automatic installation of sofware and writing a user-friendly
tool for controlling what has to be started at boot time.