I've encounter some interesting comments lately that
are remotely related to assembly programming. The first
is in this months "Linux Format" magazine. I'm now
reading "Linux Format" and the other Euro magazine
which started in Germany. They seem to be better
than anything available here in the US. Anyway,
the comment is:

"our favorite desktops and apps are snowballing
into bloat"

The author went on to say Microsoft will gain users
if Linux continues on the bloat path. He was mostly
referring to Open Office and some of the big desktops.

In another article they talk about game programming
in JAVA and how it now has good performance. ???

These comments may indicate that apps have begun to
mature and the focus change from features to performance.
I sure hope so.

-----

While browsing the Linux Assembly FAQ I found several
comments that seemed questionable. One would expect
the comments to be favorable to assembly, but the opposite
is what I found. For example this is a quote:

If any other program on your computer uses the libc,
then duplicating the libc code will actually
wastes memory, not saves it.

The above comment was listed as a reason to avoid assembly.
Is it true?

I did a few tests and found that for most programs the
opposite is true. Most static linked assembly programs
are still smaller than a dynamically linked "c" program.
The reasons for this may be:
1. The ELF format has extra sections to support dynamic
linking.
2. General purpose libraries have so much error checking
and general case code, they tend to be bloated. Even
libc. Some of this is passed on in the calling sequence.
3. Assembly code has a different mind set and the design
is often different. This results in much smaller
programs and different calling interfaces.
4. Many applications have special libraries that are not
shared. The general libraries are huge and none
of them come close to the small size of assembly programs.
This needs to be included in the overall program size.

The overall tone of the FAQ is to avoid assembly except for
optimization. They did expand on this somewhat with:

In short, main practical (need) reasons of diving into the assembly
realm are small code and libc independence.
Impractical (want), and the most often reason is being just an old
crazy hacker, who has twenty years old habit of doing everything
in assembly language.

OK, there it is. Is it accurate?
What about speed?
What about small embedded systems where hardware costs are
the key to sales?
What about bit intensive simulations?
What about OS code or Video drivers?
on and on...

Maybe, I'm one of those crazy old hackers, and maybe the
FAQ writers are missing some experience? <grin>

all the best, jeff