httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
Subject Re: OS/390 specific directory in 2.0
Date Wed, 22 Mar 2000 21:36:39 GMT
In a recent note, Jeff Trawick said:

> Date: Wed, 22 Mar 2000 15:50:09 -0500
> As of this moment, Greg A. has updated for 2.0 and if we
> can agree on where it needs to live then we can go ahead and put it
> there and not bother with separate distribution mechanisms.  Before
> long there may be other scripts required and there is OS/390-specific
> code in 1.3 that I would have thought we'd start off with.  Take a
> look at apache-1.3/src/os/os390.
I haven't studied this.  Why can't os390 coexist with unix?  There
isn't a separate src/os/linux, nor src/os/svr4, nor src/os/solaris,

> Separately there's the issue of three separate EBCDIC tables (much of
> what lives in src/os/(os390|tpf|bs2000) and the problem that, at least
> for OS/390, no one table makes everybody happy.
I believe that TPF and OS/390 both use IBM-1047 (with the LF/NL tweak),
at least nearly enough to make the compiler happy.  Beyond that, it
doesn't matter, as long as the socket front end precisely undoes the
translation done by archive extraction.

Then there's the matter of local variants of EBCDIC which Ovies had
started to ponder.  Ugh.  Rexx script to run the entire character
set through iconv and generate the translation table at compilation
time (best as close to archive extraction as possible)?  Ugh.
Use __atoe() to dynamically generate the translation at runtime?
Ugh.  (What if someone changes LOCALE between extraction and 
compilation and execution?)

Why does EBCDIC need to worry about local code pages, whereas
ASCII doesn't?  Once you accept translation, you open Pandora's
box, and everyone wants a different translation.

-- gil

View raw message