| |||||||||||
|
| |||||||||||
Thomas Bushnell removed from Hurd developmentThomas Bushnell removed from Hurd developmentPosted Nov 21, 2003 14:29 UTC (Fri) by MLKahnt (subscriber, #6642)In reply to: Thomas Bushnell removed from Hurd development by argent Parent article: Thomas Bushnell is no longer Hurd maintainer There are debates about how "microkernel" Darwin/OSX is due to a significant layer of monolithic code sitting on top of Mach. That said, I suspect that the HURD could also be an example that Theory always runs faster than Code - to implement the concepts and structures of the HURD, sufficient overhead is essential that the resulting system is not comparably paced to a monolithic kernel. The HURD has been implemented with the GNU Mach kernel - my understanding being that it is effectively comparable to the kernel at the core of Darwin and NeXTStep, but while it was sufficiently fast enough for them with their monolithic support packages, it is only capable of "Proof of Concept" performance for the HURD. Now, the as yet still under development L4 kernel is seen as the "next great hope" to make performance improvements to make the HURD viable. The need to switch between processes and in and out of system mode is that the architecture implies, however, does impact the potential performance. My experience of the HURD is that, unlike most other systems where you have the kernel, and the systems (libc, X11, GNU, etc.) being ported to it, everything is being reworked to the concepts of the HURD's design - largely all at once without sufficient minds and bodies available for the scale of such a project on a timely basis. While the concepts stay relatively stable, the microkernel(s), libraries and other parts are changing and are regularly needing to be re-ported, while certain obvious problems that would not have been left unaddressed this long on other production o/s platforms continue to linger. With that arises an impression among some of lack of progress, and yes, net progress is difficult to attain in the HURD's current context. I have argued that possibly what is needed is to focus on the core, and leave the niceties (eg. X11) until a viable base platform is ready, but let's be honest - in any volunteer project, the code produced is what interests and intrigues the programmer writing it, or is absolutely necessary to make the code he or she wants to implement work.
(Log in to post comments)
|
|||||||||||