[tex-live] TeXLive-CD/DVD (Installation)

George N. White III gnwiii at gmail.com
Wed May 23 13:55:37 CEST 2007


On 5/22/07, Edd Barrett <vext01 at gmail.com> wrote:

> On 22/05/07, Reinhard Kotucha <reinhard.kotucha at web.de> wrote:
> > We need, at least for UNIX, an installer which works on all platforms.
> > Currently it is a shell script which uses some autoconf magic for
> > platform detection (config.guess).  The shell script we have now is
> > quite fine.  I'm very happy with it and for UNIX it is definitely the
> > best solution.
>
> Hi there,
>
> I have do disagree. But allow me to justify myself.
>
> There are too many shells with different syntax. Yes, your installer
> works great on systems with bash, mostly linux i suppose. The
> installer does not work on my platform of choice. I suspect it is the
> shell (ksh), but it could be the tar implementation (not gtar). Either
> way shell commands are a bad way to go in my opinion, as they differ
> between platfroms, and windows has no good shell unless cygwin is
> installed (which is too much to ask for a texlive install I think).
> Why not use library calls from say python? This gaurantee's platform
> independancy surely?

Older systems may not have python, or may have old versions, or the
versions may have been built with features disabled.   With any scripting
language, you have to be careful to avoid  constructs specific to the version
you happen to use, or, in a few cases, supply different coding based on
the version.

If you look at the tetex scripts, say texconfig, you can find a number of weird
and wonderful hacks to support different systems.  There is a delicate
balance between taking advantage of really helpful capabilities things like
shell functions and preserving the ability to run on legacy machines.  Another
issue is that the interfaces to some utilities (sort, tail) have
changed over time.

There is no magic bullet, but there also needs to be a limit the
effort required
from developers support TL on legacy systems.

The fact that teTeX worked across the wide variety of unix platforms that were
current in the 1990's may have created unrealistic expectations that TL should
also work on the few surviving legacy unix machines.  One of the reasons those
legacy machines are still around is the difficulty of migrating some legacy
applications onto current platforms.   Today it is more important that
TL support
formatting the legacy documents created on those machines using current
mainstream platforms.

I don't see any pressing reason to discard the existing *x scripts.
You can even
use them on *x to create a tree that has Windows binaries.  I don't
have much insight
into the DVD install issue because most of my installs have used the inst .iso
mounted thru the loop device.   I do think get the impression that
installation speed
and convenience is much more important for Windows because people expect to
reinstall early and often.   A 5 hour install is one thing if you
expect to use it for the
next 18 months, entirely another if you expect to repeat it 18 times.

I do have experience writing data DVD's on Win32, where the AV
software seems to
need all day or overnight to finish scanning the input files (usually
bzip2 compressed
data files created on those legacy *x systems).  It seems to me that
some research
is needed to understand the interaction between disk layout, AV
scanning, and I/O
thruput on Windows.  This might make a nice high-school science fair project.

-- 
George N. White III <aa056 at chebucto.ns.ca>
Head of St. Margarets Bay, Nova Scotia


More information about the tex-live mailing list