> From: Jerry Feldman <[EMAIL PROTECTED]>
> Subject: Re: Linux binary compatability (was: Redhat 7.0 (xinetd))
On Mon, 07 Aug 2000, Jerry Feldman <[EMAIL PROTECTED]> wrote:
>
> I certainly agree with Derek on the binary compatibility. Maintaining
> binary compatibility causes the growth of libraries because of legacy
> code. For instance, time_t changes to 64 bits. WIth binary
> compatibility there must be a mechanism for older programs to run
> correctly without recompilation, but newer programs should
> automatically get the 64 bit time. The same goes for some data
> structures, and some major library functions and system calls. Result, bloat.
So the rapid feature bloat we see is OK but compatibility bloat isn't?
Here's a proposal: let's recycle 3% of the feature bloat and reserve
it for compatibility bloat. ;-)
Another (overly?) simplistic model of mine:
# of library developers << # of application developers << # of end-users
with the rule whenever you find yourself to the left of a "<<" you spend
the extra time and thoughtfulness to try not break people to the right.
This is ultruistic, but saves overall in people-time.
Note that when Linux started (or when just about any OS starts) the 3
numbers are actually roughly the same instead of "<<"! Everybody is
rebuilding the world nearly every week, as Ben pointed out, and
compatibility (binary or otherwise) just isn't a big issue. "Oh, just fix
it, here's how..."
But when we have # of Linux end-users = 50 million (as we seem to be
headed for) compatibility plays a larger role. Then a library developer
can't easily say something like "Oh it was dumb of me to do that struct
that old way; the new way is so much better. You 50M end-users had best
rebuild all of your apps. Oh, you have 150 in-house apps that no longer
work? Write a shell script..." Not exactly mission critical.
I don't mean to imply that the Linux and OSS folk don't care about
compatibility at all. They do. They just get a C- in it rather than the
normal A, A+ 's they get in other areas. Can't blame them,
compatibility is a boring subject ;-)
Since Linux isn't really tied to sales of hardware (unlike Tru64+alpha,
Solaris+sparc, HP/UX+PArisc... are) maybe compatibility will never play
a big role. But I like to think as the ratios of the above numbers
get larger, more compatibility will be a necessity. Maybe it's a
"life cycle" thing. We'll see how (or if) the distros handle it all.
Karl Runge
**********************************************************
To unsubscribe from this list, send mail to
[EMAIL PROTECTED] with the following text in the
*body* (*not* the subject line) of the letter:
unsubscribe gnhlug
**********************************************************
