[Sugar-devel] [Marketing] press release opportunity...

[Sebastian Dziallas]

Caroline Meeks wrote:
> On Wed, Jul 29, 2009 at 9:14 PM, Gary C Martin <gary at garycmartin.com
> <mailto:gary at garycmartin.com>> wrote:
>
>     On 29 Jul 2009, at 21:35, Walter Bender wrote:
>
>         Begin handwaving.
>
>         LiveUSB came from the world of LiveCD and with it came an "overlay"
>         concept to enable writing in what had been a read-only world. It is
>         not clear that the approach was intended for more than demonstration
>         purposes, in order to show off the power of Fedora Linux. That would
>         suggest that in the long run, we may need to revisit the way in
>         which
>         we manage user data on our images.
>
>         End handwaving.
>
>
>     +1
>
>     My gut feeling is we don't want a LiveUSB, we want a bootable USB
>     with a regular install on it. Ideally being installed from a LiveCD,
>     that can either directly boot and demo Sugar, install to a USB
>     stick, or install to a hard-disk. Once booted we'd want the minimum
>     of file writes to maximise a stick lifetime, and reduce the chance
>     of a write landing as a child unplugs.
>
>     Regards,
>     --Gary
>
>
> +1 except I think that we need it sooner not later.
>
> It is the most likely suspect on most of our stick failures. We will
> have upset teachers and kids if its not more reliable plus added expense
> and time costs.
>
> It is a blocker on:
>
>     * Reading things you've created on your Sugar Stick on a Windows or
>       Mac machine.
>     * Createing a VM that can switch stick based users without rebooting
>       out of the native OS- This will help usability quite a bit on the
>       Mac Laptops the GPA will be using next year.
>
> I'm going to try to create a spec and publicize our need for help to my
> network. I'd love help with both parts of that.

I'll throw my two cents in here, too.

I agree with Walter that we might need to revisit the whole concept in 
the long term. However, it's probably the best we can get right now.

Let me put it this way: Looking at my recent composes for SoaS, those 
were around 390 MB. This contains the compressed squashfs image. Because 
of this compression, it's read-only, but it's also that small.

Now in comparison, we could obviously place the whole file tree on a USB 
key and hack up some magic to make it boot. In fact, that's from what I 
see already the somehow preferred way used for the XO.

But for this, we'd also need to have the file tree uncompressed (since 
otherwise it would be read-only again). And that could become a problem. 
The compression works rather well for us, so if we'd try to go this way, 
we'd definitely need to move the USB key size requirement up (at least 2 
GB, if not even more).

And then, I'm not really sure if this solves the data corruption issue 
(which I haven't experienced myself, so far) - because files could get 
destroyed if the USB key is improperly removed anyway.

Caroline, maybe you could explain the way you're using to make these 
keys, because I've lost track about what the current way is.

Regarding reading contents one created in Sugar on Windows / Mac, I 
think this is still quite some time away. In fact, I'm wondering whether 
this isn't a datastore related feature. /me thinks about this...

Cheers,
--Sebastian