Re: Firefox profiles on a tmpfs (ramdisk)?

2010-10-24 Thread Phillip Susi
On 10/24/2010 10:12 AM, Jouko Orava wrote:
> The question is, are Ubuntu developers interested in this?

No.

Have you tried using libeatmydata?  The only key difference there should 
be with using tmpfs is that it effectively makes fsync() a noop.  Aside 
from fsync and friends, the filesystem cache should make using the files 
normally just as fast, but firefox makes excessive use of fsync, which 
tends to slow things down.  You should get similar performance from 
disabling the fsync calls to the normal on disk files, which you can do 
with libeatmydata.  Unfortunately Mozilla views the syncing as vital to 
prevent the profile from being corrupted, and will not work to remove them.

-- 
Ubuntu-devel-discuss mailing list
Ubuntu-devel-discuss@lists.ubuntu.com
Modify settings or unsubscribe at: 
https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss


Re: Welcome to the "Ubuntu-devel-discuss" mailing list

2010-10-24 Thread Phillip Susi
On 10/24/2010 05:31 AM, León Asad Castillejos wrote:
> Hi all, I'm León.
>
> I'm helping a friend to update to Karmic over IM. I've noticed a few... not
> bugs, but little problems.

You should compose a new message and set the subject to something 
appropriate instead of relying to the automated welcome message.

-- 
Ubuntu-devel-discuss mailing list
Ubuntu-devel-discuss@lists.ubuntu.com
Modify settings or unsubscribe at: 
https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss


Firefox profiles on a tmpfs (ramdisk)?

2010-10-24 Thread Jouko Orava
Hello,

For a couple of years now I've kept my active Firefox profiles on a tmpfs
(mounted size=128M,mode=01777,atime,auto,rw,nodev,noexec,nosuid)
on both my desktop workstation and on my minilaptop. It feels subjectively
much snappier, but I haven't actually measured the difference.
On the laptop, it also reduces disk usage, which is very nice.

The profiles are unpacked from a tarball when firefox starts, and tarred
back when firefox closes, using two script hooks in run-mozilla.sh:
[ -x /usr/lib/firefox-profiles.sh ] && /usr/lib/firefox-profiles.sh load
just before the final if clause, and
[ -x /usr/lib/firefox-profiles.sh ] && /usr/lib/firefox-profiles.sh save
just after the final if clause.

The script itself is very careful to not fail (and not to overwrite the
old tarball unless the new one is ok), and has never failed yet,
even when Firefox crashes.

I tried to get something similar upstream earlier:
https://bugzilla.mozilla.org/show_bug.cgi?id=576707
but it was rejected for complexity reasons.

If the Ubuntu Mozilla team were to add the script hook into Firefox
-- noting that it does not change the behaviour by default, only if the
hooked script actually exists --, I could create an Ubuntu package,
which when installed creates the necessary tmpfs, profile tarballs,
and modifications to the profile configs to limit the profile size
(so that the tmpfs does not grow too large).
All of the changes are also safely undoable if the package is removed.
Each user can also choose which profiles, if any, are kept on the tmpfs.

The question is, are Ubuntu developers interested in this?

It does add changes to Firefox run-mozilla.sh script, which has to be
maintained. (However, run-mozilla.sh does not seem to change often.)

If there is interest for this, I could upload or e-mail the scripts
(/usr/lib/firefox-profiles.sh for Firefox, and proposed maintainer
 scripts for a package setting up the tmpfs and moving Firefox
 profiles to a tmpfs.)

(There are also some user interface related issues to think about, too.
 For example, should the package move all profiles to tmpfs by default,
 or should it have a GUI prompting the user on the next Firefox startup?
 Is there need for a migration tool, GUI or commandline, so that users
 could choose afterwards?)

At the minimum, adding the script hook into run-mozilla.sh makes
the testing much easier. Now, I need to modify run-mozilla.sh whenever
Firefox is updated, which is a bit of a chore..

I'd be very happy to provide further information for those interested.

Cheers,
Jouko Orava

-- 
Ubuntu-devel-discuss mailing list
Ubuntu-devel-discuss@lists.ubuntu.com
Modify settings or unsubscribe at: 
https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss


Re: Welcome to the "Ubuntu-devel-discuss" mailing list

2010-10-24 Thread León Asad Castillejos
Hi all, I'm León.

I'm helping a friend to update to Karmic over IM. I've noticed a few... not
bugs, but little problems.

For example, the update manager for him showed at first 8 minutes left. Then
it changed to 34, then to 9 again. In my opinion, the timing should be
calculated having in mind the lowest KB/s ever received in the upgrade
process, and then, slowly, cut off time if the network keeps stable and
reaches high speeds. I think that the user shouldn't have "remaining time"
displayed, but "maximum remaining time" instead.

For example, if I'm in a hurry and only have 1 hour left with the computer,
and I want to know if the process will complete, I prefer it to tell me it
will take 1 hour and 5 minutes if it's gonna take 55 minutes, because there
can be a lot of problems in the downloading process. Problems like unstable
WiFi, stolen WiFi, messy cables, etc.

Second. When the update manager detects a new version, don't just display a
passive banner. It would be better if a OSD notification was displayed,
showing the update manager icon. This notification would be displayed once
or twice a week for 4 weeks, after that time, everything notifying of a new
version would be disabled, while the end user could still update manually.

I'm afraid i haven't been as clear as I wanted to be. If you need me to
clear something out, please reply and I will try to explain.
-- 
Ubuntu-devel-discuss mailing list
Ubuntu-devel-discuss@lists.ubuntu.com
Modify settings or unsubscribe at: 
https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss