Debian Bug report logs - #8471
Size and InstalledSize

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: "Brian C. White" <bcwhite@verisim.com>

To: Debian Bugs <submit@bugs.debian.org>

Subject: Size and InstalledSize

Date: Wed, 02 Apr 1997 15:53:47 -0500

Package: dpkg-dev
Version: 1.4.0.8

The "dpkg-gencontrol" program gives the "InstalledSize" in kilobytes where
the "Size" field is in bytes.  Would it be worth multiplying the installed
size by 1024 just to be consistant?

                                          Brian
                                 ( bcwhite@verisim.com )

-------------------------------------------------------------------------------
 Generated by Signify v1.01.  For this and more, visit http://www.verisim.com/

Message #10 received at 8471@bugs.debian.org (full text, mbox, reply):

From: Rob Browning <osiris@cs.utexas.edu>

To: "Brian C. White" <bcwhite@verisim.com>

Cc: 8471@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: 02 Apr 1997 15:40:27 -0600

"Brian C. White" <bcwhite@verisim.com> writes:

> The "dpkg-gencontrol" program gives the "InstalledSize" in kilobytes where
> the "Size" field is in bytes.  Would it be worth multiplying the installed
> size by 1024 just to be consistant?

Or better yet, just put the actual number of bytes.  There's no real
point in rounding, and there might be a time where those fractional
values would be useful...

-- 
Rob

Message #15 received at 8471@bugs.debian.org (full text, mbox, reply):

From: "Brian C. White" <bcwhite@verisim.com>

To: Rob Browning <osiris@cs.utexas.edu>

Cc: 8471@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: Wed, 02 Apr 1997 16:48:28 -0500

> > The "dpkg-gencontrol" program gives the "InstalledSize" in kilobytes where
> > the "Size" field is in bytes.  Would it be worth multiplying the installed
> > size by 1024 just to be consistant?
> 
> Or better yet, just put the actual number of bytes.  There's no real
> point in rounding, and there might be a time where those fractional
> values would be useful...

True, but unfortunately, the size is calculated by running 'du' which only
returns a block count.

                                          Brian
                                 ( bcwhite@verisim.com )

-------------------------------------------------------------------------------
    In theory, theory and practice are the same.  In practice, they're not.

Message #20 received at 8471@bugs.debian.org (full text, mbox, reply):

From: Rob Browning <osiris@cs.utexas.edu>

To: "Brian C. White" <bcwhite@verisim.com>

Cc: 8471@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: 02 Apr 1997 16:49:37 -0600

"Brian C. White" <bcwhite@verisim.com> writes:

> True, but unfortunately, the size is calculated by running 'du' which only
> returns a block count.

How about "du --bytes" then?

-- 
Rob

Message #25 received at 8471@bugs.debian.org (full text, mbox, reply):

From: David Engel <david@sw.ods.com>

To: "Brian C. White" <bcwhite@verisim.com>, 8471@bugs.debian.org

Cc: Rob Browning <osiris@cs.utexas.edu>

Subject: Re: Bug#8471: Size and InstalledSize

Date: Wed, 2 Apr 1997 17:22:43 -0600

On Apr 2, Brian C. White wrote
> True, but unfortunately, the size is calculated by running 'du' which only
> returns a block count.

du returns bytes if you use the "--bytes" option.

David
-- 
David Engel                        ODS Networks
david@sw.ods.com                   1001 E. Arapaho Road
(972) 234-6400                     Richardson, TX  75081

Message #30 received at 8471@bugs.debian.org (full text, mbox, reply):

From: "Brian C. White" <bcwhite@verisim.com>

To: Rob Browning <osiris@cs.utexas.edu>

Cc: 8471@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: Thu, 03 Apr 1997 11:17:24 -0500

> > True, but unfortunately, the size is calculated by running 'du' which only
> > returns a block count.
> 
> How about "du --bytes" then?

Oh, hey!  I didn't even know about that.  That would be a better solution.

                                          Brian
                                 ( bcwhite@verisim.com )

-------------------------------------------------------------------------------
     measure with micrometer, mark with chalk, cut with axe, hope like hell

Message #35 received at 8471@bugs.debian.org (full text, mbox, reply):

From: Martin Alonso Soto Jacome <masoto@uniandes.edu.co>

To: Rob Browning <osiris@cs.utexas.edu>, 8471@bugs.debian.org

Cc: "Brian C. White" <bcwhite@verisim.com>

Subject: Re: Bug#8471: Size and InstalledSize

Date: Thu, 03 Apr 1997 14:31:42 -0500

Rob Browning writes:
> Or better yet, just put the actual number of bytes.  There's no real
> point in rounding, and there might be a time where those fractional
> values would be useful...

Please don't.  The real amount of space used by an installed package is a 
number of blocks, not bytes (the file system allocates blocks after all).  
IMO, using du to measure it is the right thing to do.

M. S.

------------
Martin A. Soto J.                           Profesor
Departamento de Ingenieria de Sistemas y Computacion
Universidad de los Andes      masoto@uniandes.edu.co

Message #40 received at 8471@bugs.debian.org (full text, mbox, reply):

From: "Brian C. White" <bcwhite@verisim.com>

To: Martin Alonso Soto Jacome <masoto@uniandes.edu.co>

Cc: Rob Browning <osiris@cs.utexas.edu>, 8471@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: Thu, 03 Apr 1997 16:05:22 -0500

> > Or better yet, just put the actual number of bytes.  There's no real
> > point in rounding, and there might be a time where those fractional
> > values would be useful...
> 
> Please don't.  The real amount of space used by an installed package is a
> number of blocks, not bytes (the file system allocates blocks after all).
> IMO, using du to measure it is the right thing to do.

Just as long as it gets converted to bytes so the two fields ("Size"
and "InstalledSize") have consistant units.

                                          Brian
                                 ( bcwhite@verisim.com )

-------------------------------------------------------------------------------
 Generated by Signify v1.01.  For this and more, visit http://www.verisim.com/

Message #45 received at 8471@bugs.debian.org (full text, mbox, reply):

From: Rob Browning <osiris@cs.utexas.edu>

To: Martin Alonso Soto Jacome <masoto@uniandes.edu.co>

Cc: 8471@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: 03 Apr 1997 15:11:46 -0600

Martin Alonso Soto Jacome <masoto@uniandes.edu.co> writes:

> Please don't.  The real amount of space used by an installed package is a 
> number of blocks, not bytes (the file system allocates blocks after all).  
> IMO, using du to measure it is the right thing to do.

Do all linux file systems use the same size blocks?

-- 
Rob

Message #50 received at 8471@bugs.debian.org (full text, mbox, reply):

From: Rob Browning <osiris@cs.utexas.edu>

To: Martin Alonso Soto Jacome <masoto@uniandes.edu.co>

Cc: 8471@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: 03 Apr 1997 15:16:27 -0600

Martin Alonso Soto Jacome <masoto@uniandes.edu.co> writes:

> Please don't.  The real amount of space used by an installed package is a 
> number of blocks, not bytes (the file system allocates blocks after all).  
> IMO, using du to measure it is the right thing to do.

And I just noticed that even du's idea of what a block's size is is
dependent on the setting of an environment variable (POSIXLY_CORRECT).

Bytes are unambiguous, and accurate.  Who knows what you might want
that extra accuracy for, but what's the harm in keeping it?

-- 
Rob

Message #55 received at 8471@bugs.debian.org (full text, mbox, reply):

From: Fabrizio Polacco <fpolacco@icenet.fi>

To: Rob Browning <osiris@cs.utexas.edu>, 8471@bugs.debian.org

Cc: Martin Alonso Soto Jacome <masoto@uniandes.edu.co>

Subject: Re: Bug#8471: Size and InstalledSize

Date: Sun, 06 Apr 1997 23:38:55 +0300

Rob Browning wrote:
> 
> And I just noticed that even du's idea of what a block's size is is
> dependent on the setting of an environment variable (POSIXLY_CORRECT).

from   du --help

  -k, --kilobytes      use 1024 blocks, not 512 despite POSIXLY_CORRECT


Fabrizio
-- 
+----------------------------------------------------------------------+
| fpolacco@icenet.fi  fpolacco@debian.org  -  Using Debian GNU/Linux ! |
| 6F7267F5 fingerprint 57 16 C4 ED C9 86 40 7B 1A 69 A1 66 EC FB D2 5E |
+----------------------------------------------------------------------+

Message #58 received at 8471-quiet@bugs.debian.org (full text, mbox, reply):

From: Ian Jackson <ian@chiark.greenend.org.uk>

To: 8471-quiet@bugs.debian.org, Debian developers list <debian-devel@lists.debian.org>

Subject: Re: Bug#8471: Size and InstalledSize

Date: Sun, 11 May 97 18:37 BST

Brian C. White writes ("Bug#8471: Size and InstalledSize"):
> Package: dpkg-dev
> Version: 1.4.0.8
> 
> The "dpkg-gencontrol" program gives the "InstalledSize" in kilobytes where
> the "Size" field is in bytes.  Would it be worth multiplying the installed
> size by 1024 just to be consistant?

We certainly shouldn't change it so that the meaning of
already-distributed packages or files changes.  If we do anything
there should be a unit on the size.

I'm not convinced it's worth it.  It's pretty obvious to a human what
the units are, and a program can be made to deal with the information
in any way we feel is appropriate.

Ian.

Message #61 received at 8471-quiet@bugs.debian.org (full text, mbox, reply):

From: Rob Browning <rlb@cs.utexas.edu>

To: Ian Jackson <ian@chiark.greenend.org.uk>

Cc: 8471-quiet@bugs.debian.org, Debian developers list <debian-devel@lists.debian.org>

Subject: Re: Bug#8471: Size and InstalledSize

Date: 11 May 1997 18:44:11 -0500

Ian Jackson <ian@chiark.greenend.org.uk> writes:

> We certainly shouldn't change it so that the meaning of
> already-distributed packages or files changes.

> If we do anything there should be a unit on the size.

Yes and yes.

> I'm not convinced it's worth it.  It's pretty obvious to a human what
> the units are, and a program can be made to deal with the information
> in any way we feel is appropriate.

Not that it's really a big deal, but I don't understand why the data
was truncated in the first place.  Any interface (like dselect) could
have truncated it for user interface purposes, and truncating it has
no advantage I can see.  And who knows, in certain situations, the
information might have been useful.

I can't recall, but I may have mentioned this "objection" before.

Thanks
-- 
Rob

Message #66 received at 8471-maintonly@bugs.debian.org (full text, mbox, reply):

From: Klee Dienes <klee@mit.edu>

To: Rob Browning <rlb@cs.utexas.edu>

Cc: 8471-maintonly@bugs.debian.org,

Subject: Re: Bug#8471: Size and InstalledSize

Date: 15 May 1997 23:48:54 -0400

> Not that it's really a big deal, but I don't understand why the data
> was truncated in the first place.  Any interface (like dselect) could
> have truncated it for user interface purposes, and truncating it has
>  no advantage I can see.  And who knows, in certain situations, the
> information might have been useful.

Not that I don't agree that the data shouldn't have been truncated,
but there actually wouldn't have been any more information available.

The 'InstalledSize' field specifies the "installed size" of the
package, which generally will be a multiple of 1024 anyway (the ext2fs
block size).  Just adding the sizes of each of the files would give a
more precise, but less accurate number, as it would underestimate the
size of installed packages, particlarly those with a lot of small
files.

Of course this will get us into trouble if/when packages are installed
on filesystems with a block size different than that on which the
package was built, but that's an issue I'd personally rather defer
until it actually becomes important.

Message #71 received at 8471-maintonly@bugs.debian.org (full text, mbox, reply):

From: Rob Browning <rlb@cs.utexas.edu>

To: Klee Dienes <klee@mit.edu>

Cc: 8471-maintonly@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: 15 May 1997 23:28:07 -0500

Klee Dienes <klee@mit.edu> writes:

> The 'InstalledSize' field specifies the "installed size" of the
> package, which generally will be a multiple of 1024 anyway (the ext2fs
> block size).  Just adding the sizes of each of the files would give a
> more precise, but less accurate number, as it would underestimate the
> size of installed packages, particlarly those with a lot of small
> files.

You know, I hadn't even thought of that, but I still maintain that
since filesystems can have different block sizes, the ideal thing
would be for the package to contain all the info, and then the install
program, if it matters, could query the filesystem (via proc or
whatever) and figure out what the real size on disk would be.

A very minor issue...
-- 
Rob

Message #76 received at 8471-maintonly@bugs.debian.org (full text, mbox, reply):

From: Klee Dienes <klee@mit.edu>

To: Rob Browning <rlb@cs.utexas.edu>

Cc: Klee Dienes <klee@MIT.EDU>, 8471-maintonly@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: Fri, 16 May 1997 00:51:16 -0400

> You know, I hadn't even thought of that, but I still maintain that
> since filesystems can have different block sizes, the ideal thing
> would be for the package to contain all the info, and then the install
> program, if it matters, could query the filesystem (via proc or
> whatever) and figure out what the real size on disk would be.

Agreed, but that is potentially quite a big list (consider
picon-domains), and not really something you want to put in a
'Packages' file.

One ugly way to do this is to have a new field called
'Filesize-Histogram' that contains a histogram of the number of files
in each of several categories of file sizes, where the maximum file
size increases by a power of two for each bucket.

So for example, the package:

drwxr-xr-x root/root         0 1997-03-08 15:48 ./
drwxr-xr-x root/root         0 1997-03-08 15:48 usr/
drwxr-xr-x root/root         0 1997-03-08 15:48 usr/bin/
-rwxr-xr-x root/root     11208 1997-03-08 15:48 usr/bin/xzoom
drwxr-xr-x root/root         0 1997-03-08 15:48 usr/man/
drwxr-xr-x root/root         0 1997-03-08 15:48 usr/man/man1/
-r--r--r-- root/root      2368 1997-03-08 15:48 usr/man/man1/xzoom.1x.gz
drwxr-xr-x root/root         0 1997-03-08 15:48 usr/doc/
drwxr-xr-x root/root         0 1997-03-08 15:48 usr/doc/xzoom/
-rw-r--r-- root/root       788 1997-03-08 15:48 usr/doc/xzoom/copyright
-rw-r--r-- root/root       176 1997-03-08 15:48 usr/doc/xzoom/changelog.Debian
-rw-r--r-- root/root       110 1997-03-08 15:48 usr/doc/xzoom/buildinfo.Debian
-rw-r--r-- root/root       988 1997-03-08 15:48 usr/doc/xzoom/README
drwxr-xr-x root/root         0 1997-03-08 15:48 usr/lib/
drwxr-xr-x root/root         0 1997-03-08 15:48 usr/lib/menu/
-rw-r--r-- root/root        40 1997-03-08 15:48 usr/lib/menu/xzoom

Would have the following entry:

Filesize-Histogram: 8      3      0     2       0     1      0      1
                    (dirs) (256)  (512) (1024) (2048) (4096) (8192) (16384)

Then since all filesystems generally use a power of two as the
blocksize, you can always figure out the installed size (you'll
actually be overestimating it a bit because some of the directories
will already be provided elsewhere).

This is sufficiently ugly, though, that I'd really rather just wait
and hope a better idea comes up before we do anything drastic.

Besides, even this doesn't take into account directories that take up
more than one filesystem block, and you can't compute that without
knowing which files are going to end up being installed where.

Message #81 received at 8471-maintonly@bugs.debian.org (full text, mbox, reply):

From: Rob Browning <rlb@cs.utexas.edu>

To: Klee Dienes <klee@mit.edu>

Cc: 8471-maintonly@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: 16 May 1997 00:16:46 -0500

Klee Dienes <klee@mit.edu> writes:

> Agreed, but that is potentially quite a big list (consider
> picon-domains), and not really something you want to put in a
> 'Packages' file.

OK, OK, I give :>

Wow.  I really hadn't thought that one through.  I was just blundering
along assuming that if you had one number you'd be fine, but of course
you wouldn't.  I'll shut up and let you go do something important now.
-- 
Rob

Message #86 received at 8471-maintonly@bugs.debian.org (full text, mbox, reply):

From: Klee Dienes <klee@mit.edu>

To: Rob Browning <rlb@cs.utexas.edu>

Cc: 8471-maintonly@bugs.debian.org

Subject: Re: Bug#8471: Size and InstalledSize

Date: Fri, 16 May 1997 02:32:38 -0400

> you wouldn't.  I'll shut up and let you go do something important
> now.

Something important?  Not likely.  At the moment, "sleep" is looking
like a top contender.  :-).

Message #93 received at 8471-done@bugs.debian.org (full text, mbox, reply):

From: Wichert Akkerman <wichert@cistron.nl>

To: 8471-done@bugs.debian.org

Subject: Re: Size and InstalledSize

Date: Sat, 21 Apr 2001 04:55:33 +0200

The discussion in this bugreports clearly shows that there is
no way to fix this bug, so I'm closing it.

Wichert.

-- 
  _________________________________________________________________
 /       Nothing is fool-proof to a sufficiently talented fool     \
| wichert@cistron.nl                  http://www.liacs.nl/~wichert/ |
| 1024D/2FA3BC2D 576E 100B 518D 2F16 36B0  2805 3CB8 9250 2FA3 BC2D |

Send a report that this bug log contains spam.