Tom Christiansen wrote to Dan Sugalski:
> >Pity we've seen such a slowdown in recent perls. I'm hoping efficiency
> >hacks are high on the list for 5.7/5.8
>
> I think the bloat is due to Unicode. I think the speeddowns are
> because of Larry not really being so into the release process. He
> hates slow, and if he noticed, he'd tweak or thwap it into something
> faster.
Without a doubt this to/from utf8 stuff slows things down.
To top it off the apparent goal is to treat all input as utf8 eventually.
On OS/390 one has a choice of several programming languages including:
Cobol, Fortran, C (including gcc), C++, Java, sh, Rexx, and I am pretty
sure that Tcl and Python ports are available too. Perl, as of 5.6.0, is
the only programming language (outside of maybe APL, and even that only added
glyphs to your native char set) that will attempt to dictate the coded
character set allowed for input. That is pretty tyrannical.
Not only that but utf8 input seems to violate one of Larry's own goals
with unicode support, namely allowing for the writing of your variable
names in your non-7-bit-ascii glyph set. The font factories are/will be
producing, e.g. Han glyphs for UCS-2 (i.e. unicode) _not_ for utf8 which
is only a legacy->unicode transform intermediary set (mostly for email).
I realize that native char set I/O stuff is high on the Todo list but the
utf8 in 5.6.0 comes as bad news for lots of reasons.
Peter Prymmer
