Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!elroy.jpl.nasa.gov!sdd.hp.com!decwrl!decwrl!usenet.coe.montana.edu!grapevine.lcs.mit.edu!ai-lab!life.ai.mit.edu!mycroft From: mycroft@trinity.gnu.ai.mit.edu (Charles Hannum) Newsgroups: comp.os.386bsd.questions Subject: Re: FreeBSD or NetBSD-0.9 ? Date: 25 Aug 1993 00:19:36 GMT Organization: MIT Artificial Intelligence Lab Lines: 38 Message-ID: <MYCROFT.93Aug24201936@trinity.gnu.ai.mit.edu> References: <1993Aug24.094944.15984@info.brad.ac.uk> <1993Aug24.132725.3171@sophia.smith.edu> <25dijl$ocr@pdq.coe.montana.edu> NNTP-Posting-Host: trinity.gnu.ai.mit.edu In-reply-to: osyjm@cs.montana.edu's message of 24 Aug 1993 17:19:49 GMT FreeBSD's goal is to be stable (not necessarily machine uptime), but slowly changing, as such, it will by nature not be bleeding edge. There isn't too much that is massively changed from 386bsd 0.1, with the exception of patches to make the machine run longer, and update all the utilities. This is misleading. Several key things have been changed in FreeBSD, such as the sio driver, and have to date introduced serious bugs. It would be a fallacy to say that it is a small update that is unlikely to cause any problems. Also, given the number of NetBSD changes I see being absorbed into FreeBSD, I would hardly call it `slowly changing'. NetBSD's goal is stability (+ being a research OS), however they are replacing subsystems with other (they feel better) subsytems as they come across them. I can't parse the end of the above sentence. It is true that we (NetBSD) have replaced some key components of the kernel--e.g. a new ring buffer implementation with a clist interface, a new execve(), replacing the rlist code with rmaps, etc. However, it is not at all a matter of `as [we] come across them'; it is a careful and thorough examination of the state of the system, and replacement of components which are broken or less functional than is desired. For example, the clist interface makes BSD-based drivers like PPP very easy to port; the new execve() sets protections correctly and is much easier to support multiple architectures with; rmaps are faster and probably smaller than rlists, and are applicable in more circumstances. There also is the fact that they seem to be moving to 4.4-style stuff as much as possible. (Which is not inherently a bad thing, just different). We often adopt 4.4 updates, but then so does FreeBSD.