lkml.org 
[lkml]   [1998]   [Nov]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: unremovable files and possible fs corruption (2.1.123)
On Sun, 1 Nov 1998, Brandon S. Allbery KF8NH wrote:

> In message <Pine.LNX.3.93.981101115722.597A-100000@kjahds.kjahds.com>,
> Kenneth
> Albanowski writes:
> +-----
> | > Today I can write code that avoids libc altogether and continues to
> | > work, but if libc becomes the published interface, that guarantee is
> | > gone.
> |
> | Also, in my interpreter-evangelist hat, I'd like to be able to bind
> | anything to anything else, at any level, which at some point means
> | teaching an interpreter to talk directly to the kernel, without libc in
> | the middle. This ought to be possible.
> +--->8
>
> But that makes changes to the kernel interface difficult or impossible. It
> requires the kind of ugliness that you find in NetBIOS, or a special syscall
> that every program must make to declare which variant of system interface
> they want to use, or requires the syscall interface to expand without limit
> to accommodate changed syscalls. (And then we get screams from the folks
> who bind directly to the kernel because they can't access the new
> functionality without rewriting their programs.)

I wouldn't argue _too_ strenuously, as just about all of these things have
been tried. We've already got duplicated syscalls, syscalls that look for
"magic" numbers to figure out which syscall version is being invoked and
probably some other variations.

But that isn't really what I was talking about, in any case. Sure,
anything using the syscalls directly needs to know what it is doing. The
issue is that the more magic libc uses to talk to the kernel, the less
published the kernel syscalls become. Yes, libc certainly can and should
do what it wants to make the C world a better place, but that shouldn't
detract from anything else using the kernel syscalls directly.

As of 2.0.x, this was already getting messy. (Especially around the
duplicated syscalls, and the '#ifdef ALPHA' bits.) I've not look at 2.1.x
closely enough to see whether there has been improvement.

--
Kenneth Albanowski (kjahds@kjahds.com, CIS: 70705,126)




-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:45    [W:0.048 / U:0.024 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site