[tex-live] TL2004: Technical problems and testing

[Olaf Weber]

Karl Berry writes:

>     Basically, with plain TeX mode the 8bit stuff should be off.  With
>     latex, my understanding is that people want it on, and same for
>     context.  

> Ohhh, this is yet another wrinkle.  I am not sure at all that it should
> depend on the format.  I haven't heard that proposed before.

It's part of the reason why xord/xchr/xprn are dumped in the format.
Remember that the trip test requires the original "printable ASCII"
behaviour, and that this is the documented behaviour for plain TeX as
well.  And writing to terminal and log are specifically exempted from
this requirement, though the trip test is best run in the C locale at
present.

> We need more input. Folks, shall we:

> 1) keep tty/log output locale-dependent via isprint, as it's always been
>    before?  Users and systems then set LC_CTYPE (or whatever) to control it.
>    Or:   

> 2) make -8bit + tcx files affect tty/log output, even when LANG=C?
>    Would this perceived to be a general improvement for users? 
>    And:

> 2a) If 2, should we make all the formats the same by default, or turn
>    off 8bit tty/log for plain, or only turn on 8bit tty/log for latex?
>    (I'm guessing context would also want 8bit tty/log, if latex does.)

> Ultimately, I guess I personally don't have strong feelings about it.
> It seems desirable to me for all formats to be treated the same.

>     You'd have to write a 7bit.tcx that actively suppresses printability
>     for all but the range 32..127 inclusive.  Easy to to with the new
>     code, but it would have to be done.

> Understood.

>     There are currently _no_ settings for this in texmf.cnf.

> Feasible to add, or not?

> I'm imagining something like
> translate_file = 7bit.tcx
> ...
> (We wouldn't have any setting in there by default.)

I suspect the damage an ill-chosen default could cause would be
considerable.

-- 
Olaf Weber

               (This space left blank for technical reasons.)