ISO size breakdown? ISO diff?

[giovanni_re]

Great work, Amedee :)

Correcting the one little mistake you made should enable you to see a _big_ difference, I think. :)   See below.


On Wed, 20 Jan 2010 14:11:25 +0100 (CET), "Amedee Van Gasse (ub)" <amedee-ubuntu at amedee.be> said:
> On Wed, January 20, 2010 09:49, Jonas Norlander wrote:
> > 2010/1/20 Amedee Van Gasse <amedee-ubuntu at amedee.be>:
> >> On 18-01-10 09:53, giovanni_re wrote:
> 
> >>> Also, is it possible to diff the different ISOs, so that if you have
> >>> one
> >>> already,&  want one other, you don't have to dl all 700MB, just the
> >>> diff?  Like, between 9.10 Ub&  KUb for x86?
> >>
> >> No.
> >
> > I think you are wrong. There are programs that can diff binary files,
> > for example xdelta. Is this not what he ask for? I haven't used xdelta
> > and have no clue how it will work on ISO files.
> 
> It is an exercise in futility.

Good work, Amedee - I think it is an exercise in progress, when done right. You've made a good first step here.  You have the process right, but the files wrong.  just use the proper files, & please tell us the redone #'s, and I think there will be a big improvement. :)


> 
> ----------CUT----------
> amedee at intrepid:~$ sudo aptitude install xdelta
...

> amedee at intrepid:~$ cd download/
> amedee at intrepid:~/download$ wget
> http://ftp.heanet.ie/pub/ubuntu-releases/karmic/ubuntu-9.10-desktop-i386.iso
> http://ftp.heanet.ie/pub/ubuntu-releases/jaunty/ubuntu-9.04-desktop-i386.iso
> --2010-01-20 13:42:17-- 
> http://ftp.heanet.ie/pub/ubuntu-releases/karmic/ubuntu-9.10-desktop-i386.iso

Er, no Amedee - not between #'d _versions_, but between, Ub & KUb ISOs for the _same_ version: ex, 9.10 U & KU ISOs.  

Basically, the only difference should be the GUI. What % of the ISO is the GUI? 5%? 10%? 20%

Point: If a person already has one ISO, they should be able to just get the diff to the other, & that should (I'm guessing) be only about 10% of 700MB - a BIG savings in download MB, especially for slower net connections.


===== NEXT STEP FORWARD: PARTIAL ISO DOWNLOAD & COMBINE SYSTEM
Moreover, to take this idea to its next productive level, one could have a _system_ that enables creating different sized ISOs - to save dl & install times.

Ex: 
1) Basic ISO parts: kernel, shell, networking, basic stuff. Maybe 200MB
2) GUI This might be 200MB.
3) Application sw: this might be 300 MB

Then, you have tools to dl & put together an ISO from these 3 parts.  Everyone dls the Basic set.  Everyone dl's _one_ of the 2 GUI sets: G or K.  Optionally, people dl the applications.

This would reduce ISO dl size totals.  It saves data transmission, & is more convenient for slow net connection users. :)


...
> amedee at intrepid:~/download$ xdelta delta ubuntu-9.04-desktop-i386.iso
> ubuntu-9.10-desktop-i386.iso jaunty-karmic.delta
> amedee at intrepid:~/download$ ls -hl *.iso *.delta
> -rw-r--r-- 1 amedee amedee 576M jan 20 13:50 jaunty-karmic.delta
> -rw-r--r-- 1 amedee amedee 699M apr 20  2009 ubuntu-9.04-desktop-i386.iso
> -rw-r--r-- 1 amedee amedee 690M okt 28 22:14 ubuntu-9.10-desktop-i386.iso
> ----------CUT----------
> 
> Using a delta for the ISO files would mean a 17% decrease in network
> traffic (from 690MB to 576MB) 

No, see above - you are doing 904 to 910.  I'm talking about 910 (or 10.04 ;) ) Ub & KUb ISOs. ;)  :)

Leaving the following here, for reference, so you can see, for comparison, how much less dl there is, when you do the #'s for the question I asked about. ;)


> but a 185% increase in storage (from 690 MB
> to 1965MB), because you would have to keep the old ISO before you can
> apply the path. And Canonical would have to store them on their servers
> too. And they would have to replicate all the diffs to all their mirrors,
> so that would mean a HUGE increase in network traffic. Plus you would
> need
> checksums for all the diffs. A lot can go wrong in transport.
> 
> In other words, the answer is no. Not because is technically impossible,
> but because it isn't practical and hardly worth the effort. As we say in
> Dutch, er zijn nog andere katten te geselen. (There are other cats to
> whip
> = there are more important things to do)

Great work A. :)  Move past the one little mistake, of not 904 vs 910, but 910 U vs KU, recalculate the #'s, & I think you'll see a _big_ savings for download sizes. :)

I'm looking forward to learn what new #'s you get. :)


> 
> -- 
> Amedee Van Gasse