Re: kawa verification questions / was: Best way to produce .exe files for Windows Kawa apps (fwd)

[David Chase <chase at world dot std dot com>]

At 06:26 PM 4/26/2002 -0700, Per Bothner wrote:
>>  kawa-1.6.97.=build.gnu.text.PadFormat
>
>Ouch.  This is because the Makefile rule ...

Hooray.  We get a lot of that, from outfits that
are much larger and well-known than the Kawa effort.
No names, of course :-).  You should be happy to hear
that aside from that and the javax.servlets mess, there
were no other warning messages.

>Well, using servlets is optional.  I figure if you're
>interested in using kawa servlets, you'll know how to
>get a copy of the servlets classes.

>Probably the cleanest solution is to not include the classes
>that depend on servlets in kawa.jar - on the other hand having
>them there is convenient.  I don't know what's best.

I don't have any gripe with them being included; it is more
or less what Java is all about to have optional packages, dynamic
loading, that sort of thing.  Long, long ago, we whacked our
compiler/linker to be tolerant of all sorts of surprises, all
in the cause of acting "just like" a dynamic system.

The real problem is more that there's at least two places to get
something that claims to be javax.servlet, and they aren't
necessarily compatible (maybe the one that is distributed
with Volano is very old).  In the installation page, you
should probably mention that "if you want to use kawa
servlets, you will need the classes in [ insert pointer to
compatible version here ]".

Think of it as greasing the skids for new/casual users.
No bad surprises, no maze-of-twisty-passages
documentation.

An unrelated question -- when I define a function at
the console level, is it just turned into bytecodes
that are then JITted or interpreted?  It looked like
that to me, I just wanted to check.  I assume, if I
eyeball the documentation, that I'll find out how
to compile things to bytecodes AOT if I want to do
that.

David Chase