LWN.net Logo

name space ideas

name space ideas

Posted May 4, 2008 3:03 UTC (Sun) by jd (guest, #26381)
In reply to: name space ideas by ccyoung
Parent article: On the conviction of Hans Reiser

I personally think the idea of programmatic, polymorphic namespaces to be interesting. How you
would actually make use of such a design is another matter. What, really, does a polymorphic
namespace mean at the conceptual level? If there is no concept behind the technology, then all
you have is a bunch of instructions. (Yes, I dislike modern art, too.)

You also have to consider the overheads. Modern hardware is serial and procedural. There's a
huge overhead in translating between modern programming methods and what a computer actually
does as it is. Adding yet more concepts alien to the underlying architecture is just going to
slow things down. It has to, because this type of abstraction is just a fancy way of emulating
in software those things the hardware cannot do, and software emulation is always going to be
slow. (Regular abstraction hides the specific details of how things are done, but doesn't
change the fundamental ideas.)

Now, for some things, the speed of the software isn't the most critical factor. Aerospace and
medical software, for example, have to have a very high degree of reliability. Speed is a
secondary consideration in such cases, and so you actually want any kind of abstraction that
reduces the chance of failure, even if it sacrifices performance to do so. You don't use
filesystems a whole lot in such mission-critical systems, and it's certainly not clear that
Reiser4 is going in a direction that would even be helpful in achieving extreme reliability
even if they did, but one could imagine that intelligent filesystems could be valuable in that
sector. Someday.


(Log in to post comments)

Copyright © 2013, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds