[Tfug] fhs on /var/cache (debian)

[John Gruenenfelder]

On Mon, Jul 24, 2006 at 01:37:41PM -0700, t takahashi wrote:
>hi john,
>
>hope you've adapted to the humidity.

Barely... the past week and a half has not been pleasant.

>to avoid further misunderstandings, my points were:
>
>1.  varcache should (imao) be a cache.
>2.  fhs says it should be a cache.
>3.  debian says that debian follows fhs.
>4.  debian does not follow fhs (i tested it).
>5.  it does #4 with reckless abandon (i tested it).
>6.  my question: has anybody besides me confirmed it empirically (i.e.
>by testing it on a whole system systematically from boot time -- not
>just deleting one apt dir on a running system)?
>
>did you think i was saying something else?  it seems like you did,
>because you are saying #1 and #2, which were *my own* points -- i even
>included the fhs definition!  (it's kind of like i say x is true, then
>you say "no, you're wrong!  x is true!")

I must have misunderstood your earlier email.

>you disagree with #4 and #5.

Correct... or at least I don't agree with them as strongly as you do.
Ignoring debconf data for now (since we both agree it is in the wrong place),
the /var/cache/apt directory seems entirely appropriate.

That dir contains only data which apt downloads during operation.  It's not
entirely transitory, otherwise it would be in /tmp I suppose.

>1.  does this mean that you also tested it the hard way (i.e. by
>booting to clean systems without /var/cache and without /var/cache
>files respectively and trying various programs) as i did and got
>different results?  if so, i'd like to know your results if you would
>like to share them.

Not the hard way, per se.  I did test it, but only in so much as I axed
/var/cache/apt and tried to use apt.  It is not an entirely hands off
procedure, though.  apt will not run if the directories are not present.  So I
had to make /var/cache/apt, apt/archives, and apt/archives/partial.  Once I
did that, apt ran fine and I could do an "apt-get update".  Also, even though
it didn't run when lacking the dir structure, the error message it gave me
told me which dir was missing.

>2.  does it mean that you believe that if dpkg works (not yet
>confirmed), then the system is not broken?  if so, i respectfully
>disagree.
>
>my definition of "the system is not broken" is that you can update and
>upgrade your debian system and get it to a completely working state
>without too much effort.  if you can't, then it is broken.
>
>you must be using a different definition, because here is your comment on #4:

Yes... my definition of broken is a little different from yours.  If the
package manager, or some other critical component, is broken, then the system
is broken.

Admittedly, apt makes things so much nicer, but the system can work without
it.  And since apt's dependencies are small, it's an annoying but not overly
difficult task to get the proper debs and install them.

>> If you didn't back it up to begin with, you can just have apt fetch it again
>> for you.
>
>not true.  i tested it the hard way.  did you?  if you got different
>results, please tell me, because it means that my testing was flawed
>or my system was in a different state.

See above.  It wasn't a hands off process, nor was it the most user friendly
process, but it did work and it didn't require fetching data from a
backup.  (again, I'm speaking only of apt here)

>in any case, thanks for replying.  perhaps you *are* trying to provide
>information that i didn't already put in my post, but it's not clear
>from your posts.

I think we had some conflicting definitions there for a bit.


-- 
--John Gruenenfelder    Research Assistant, UMass Amherst student
                        Systems Manager, MKS Imaging Technology, LLC.
Try Weasel Reader for PalmOS  --  http://gutenpalm.sf.net
"This is the most fun I've had without being drenched in the blood
of my enemies!"
        --Sam of Sam & Max