Proposal for TeXLive-n CDs

Sebastian Rahtz
Tue, 21 Mar 2000 16:33:52 +0000 (GMT)

Nelson H. F. Beebe writes:
 > Since the TeXlive-n CDs are bursting at the seams, we are never going
 > to be able to fit all of the desired binary distributions on a single
 > CD.
true. one could easily double the number of binary sets

 > However, why not put a pointer on the CD to
 > (N = 1, 2, 3, ...)  that will always
 > have the latest binary trees for each architecture.

indeed, that is precisely the way Fabrice and I are thinking with
the .tpm files ( the plan is
to have them point at multiple resources, including ones on the net.

the Win32 installer which Fabrice is working frantically on at this
moment has started the process; it offers to download some non-free
software for you at install time (eg winedt, gs6, french package), and 
clearly we can have future versions of this program do upgrades.

 > On the TeXLive-4 CD, they range in size from 9MB (i386-linux-libc5) to
 > 22MB (hppa1.1-hpux10.10), with an average of 14MB.  Thus, people could
 > transfer them over the Web, and we could maintain a permanent
 > repository for each issued CD that could include both updated trees
 > (e.g., compiled with newer compilers or higher optimization levels),
 > and new trees that weren't ready in time for the CD burn, or were not
 > widely-enough used to justify space (e.g., non-Intel Linux, FreeBSD,
 > OpenBSD, NetBSD, ...).

we do have FreeBSD this time, actually :-}

 > Indeed, FreeBSD has an interesting idea: the tree /usr/ports contains
 > in each directory a simple Makefile, and a configuration file that
 > tells where on the Internet the master source resides, but with no
 > source present in the directory.  When you do
 > % cd /usr/ports/packagename
 > % make

I can imagine making that work

I do agree that the whole install thing is the way to develop, there
are several papers about it at TUG2000....