Date: Thu, 21 Sep 1995 07:39:52 -0700 From: patl@asimov.volant.org To: pechter@shell.monmouth.com Cc: ports@freebsd.org Subject: Re: ports startup scripts Message-ID: <9509211439.AA22140@asimov.volant.org>
next in thread | raw e-mail | index | archive | help
|> > I suspect that most of the problem with the SVr4/Solaris/HP-UX startup |> > script system is poor documentation. And a lot of the people complaining |> > are really complaining about the change, not the actual result. Any |> > change we make will suffer from that, no matter how good it is. |> |> It's not documentation. See the Nemeth Sysadmin book Edition 2 (the red |> cover)... The yellow one documented the run levels with SVR2 (I think). Any documentation that doesn't come with the system is obscure. Unix users and sysadmins tend to expect to find everything they need on-line. |> Actually, it appears to be a cultural problem. Since there's no |> standard "Unix" -- there's really two -- BSD and SYSTEM V |> you get the one true Unix religious bigotry. Only two? Isn't that something like saying "There are two kinds of Christianity - Roman Catholic and Protestant" ? :-) :-) [ That was intended as a wry observation, not an invitation to rathole on varieties of unix... ] SVr4 was supposed to merge the two camps again by incorporating the advantages of both systems. It fell down a bit in the areas where both provided equivalent functionality in incompatible ways. And they -really- screwed up a few things (like serial port and printer administration.) |> I've worked with both. I've been the sysadmin on SunOS, HP-UX, Solaris 2.4, |> DC/OSx (SysVR4), OS/X (which had available both the AT&T and BSD init |> and the Sys Admin would install EITHER ONE based on preferences at the |> site). |> |> (Actually the capability to support both ways wouldn't be bad here... |> how about keeping the old BSD init method as an option) If that can be done easily and cleanly, I'd go for it. |> At Pyramid's NJ training facility we noticed the following... |> The Sys V method was pushed heavily in my classes as the method with the |> most customization... However my office ran with the BSD init -- since the |> rest of the office learned UNIX on the west coast -- while the bunch |> of folks who came out of the telcom business here (ex-AT&T and Bellcore |> folks) ran with the SysV setup. Which tends to support my point about inertia being the prime factor. |> > You make it sound like the folks working on FreeBSD would make changes |> > just to be different from SYSV. I sincerely hope that is not the case. |> > We should strive to produce the best unix-derived system that we can; |> > but vigorously fight the Not Invented Here syndrome. If somebody else |> > has a better solution than the one we are using, we should feel perfectly |> > free to adopt it. Or, if we can, improve it further. |> |> Agreed... it looks like the argument comes down to NIH and that SysV's |> startup complicates things more than the BSD /etc/rc /etc/rc.local does. I still think that complication is more apparent than real. In some ways, it has actually made things easier by making some of the decisions more obvious. (E.g., which run-level to put a link in corresponds to which major section of rc or rc.local to insert your changes into.) |> However, a new user editing rc or rc.local and screwing up can cause a lot |> of problems. I had to fix another admin's SunOS 4.1.3 machine when he |> screwed it up so bad that the shared libraries weren't mounted. Exactly. And the SVr4 method makes life -MUCH- easier for anyone building an installation package for add-on software. Scripts to safely modify rc or rc.local have to make some scary assumptions... |> I think we should go the SVR4 route and I'm willing to document it... I'll support you all the way. (I'll offer to help, but I'm not sure how much use I can be in this.) -Pat
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9509211439.AA22140>