Subject: Re: Can't build userland, resultant binaries are not executable
To: None <port-sparc64@NetBSD.org>
From: Mike Parson <mparson@bl.org>
List: port-sparc64
Date: 03/22/2005 19:15:16
On Tue, Mar 22, 2005 at 05:01:19PM -0500, Dave McGuire wrote:
> On Mar 22, 2005, at 4:23 PM, Greg Earle wrote:
>>> I guess you like logs and files in /tmp filling up your filesystem.
>>> What about what the *application* is "most comfortable" with?
>>> Operating systems aren't about administrator comfort...they're about
>>> controlling access to a computer's resources.
>>
>> Logs tend to be rotated, last I heard.  

Yeah, I rotate out this month's, but I'm a pack-rack, I have logs going
back for years.  Need to archive those to removable media one of these
days.

>> And as for them - or /tmp files (I use an MFS /tmp anyway, so that's
>> "separate" - happy?) filling up disks, I offer you this link:
>>
>> http://www.overstock.com/cgi-bin/d2.cgi?
>> PAGE=PRODUCT&PROD_ID=1106306&cid=25608&fp=F
>>
>> In these days when you can get 73 GB SCSI disks for US $80, the old
>> "But you might fill up the disk!" arguments don't hold much weight.
>
> Yeah, yeah.  The "disk is cheap so who gives a damn" argument is
> popular, I just don't like it. ;)

They've been making that argument since the early 90s.  Made me sick
then too. =)

> ...
>> I separate my 9 GB boot disk.  But the point I was trying to make is
>> that in this day and age of 73 GB SCSI drives and 400 GB EIDE drives,
>> a lot of the old arguments for partitioning have gone out the window.
>
>   But only some of them were based on limited disk space.  What about  
> FFS optimizations based on access patterns?  What about things like the  
> ability to mount /usr read-only?

And secure run-levels, mounting / & /usr ro, /var append only,
etc. Stuff one might do on a mult-user system, but probably not on
a single-user workstation.  I still build most of my systems with
seperated filesystems.  For all the reasons already mentioned in this
thread, as well as a couple of others I'm sure I'm forgetting at the
moment, I'll have to dig up one of my historical rants on the subject.

Also, backups. dump still only backs up whole filesystems.  Sure, you
could set no-dump flags on trees you don't want backed up, or I could
use something other than dump.  But I like it, it knows how to do
incrementals, scripts easily, simple enough to pull out just the files
you need, etc.

>> "Sick to your stomach"?  I suggest meditation if it stresses you that
>> much :-)
>
>   Well meditation might make me more able to deal with the idea of
> people thinking that certain things have become "OK", and by extension
> other issues that come up frequently such as the x86 architecture
> being "modern" and similar arguments, but I'd rather just try really,
> really hard to do what I think is the right thing to do.

>   Not trying to be argumentative here...just voicing an opinion.  A
> strongly-held one borne from years of experience just like you, but an
> opinion nonetheless.
>
>>  I've been running NetBSD/SPARC since 0.9 as well (early 1994, if
>> I recall correctly; with one abortive foray into NetBSD/SPARC64).
>> I don't see how it's cut-and-dried that it's "good engineering
>> practice".

I still tend to think its good practice, but there are always exceptions
to the rules.  I can't say that I've never set up a system with
everything under /, but it always pains me when I do.

>   Well der Mouse has made an excellent point on this matter, to which I  
> will reply in a moment.
>
>>>  Kids these days, indeed.
>>
>> I'm 46.  :-)
>
>   Ahh, you old fart.  I just turned 36 yesterday. ;)

32, yesterday. =)

Happy belated.

-- 
Michael Parson
mparson@bl.org