[SoaS] [Sugar-devel] SOAS 2 problems

[Bernie Innocenti]

On Fri, 2010-01-29 at 11:20 -0700, Douglas McClendon wrote:
> Inverse strawman, nicely torn down, then posited.

lol, I didn't realize I had been so clever :-)


> The reason for writable root as opposed to readonly root is convenience. 
>   There was(/is?) a fedora/redhat readonly root project for a long time, 
> I forget its name.

Stateless Linux?

An endless chase of all the obscure programs that get upset when they
cannot write a random file somewhere... it couldn't possibly have
succeeded :-)


>   The current method caught on, because warts and all, 
> people liked the results better.  Unionfs is a similar alternatve to 
> readonly root, and was also chosen enmasse by other projects.  It 
> suffers from similar, but differing warts.

Agreed.


> > But then, why waste the extra 580MB of disk space for the squashfs
> > image?
> 
> immutable installation image.

...which is supposed to make the OS more robust, not less :-)


> Bernie, I agree with the fundamental aspect of your argument, that a 
> real install to usbstick is better than a 'live' install.  But with the 
> caveat "when you have a sufficiently cheap, sufficiently large, 
> sufficiently reliable/performing stick".
>
> That caveat is become easier to get past as the months and years roll by.

Agreed here too.


> In general it has been my experience that fedora/redhat devs are far too 
> eager to blow off users with older hardware.  I understand it makes 
> design simpler, but it is entirely at odds with a project that is aiming 
> to recycle older computers in the third world.

My personal viewpoint here is that users stuck with older hardware can
keep using older software (plus their stability and security updates, of
course).

Sometimes, even ancient hardware turns out to be capable of running even
the latest and greatest software, but compromising new designs in order
to support old systems is a great recipe for market failure.

It doesn't even matter whether you're a business or a free software
community; you very quickly end in the same dark corner of Puppy Linux,
LXDE, XFCE... just to mention those that were most successful.

These projects evidently lack resources even to perform trivial bug
fixes in a timely fashion. I'm talking out of personal experience,
because I tried switching to a lightweight desktop multiple times, but
failed due to immaturity in basic usability features such as power
management, wifi networking, RANDR, dpmi, audio... even keyboard
settings!

But I'm digressing too much... If Fedora seems too much on the other
side of the obsolescence curve, it's because it chose intentionally to
focus on innovation. Though I guess it's ok, because most end-users get
to receive the fruits of this work with a delay of 1 or even 2 years. As
we're workoing on Fedora 13, most OLPC users are still using build 802,
based on Fedora 9, and will hopefully upgrade to Fedora 11 in the near
future.

(I'm all for tightening the user-developer feedback loop, whenever
feasible)


> I'd guess (I'm not speaking for anybody), that the 1G stick is still a 
> target of SoaS, and will be, for at least another year or two or three. 
>   Whereas fedora devs don't give a crap about that use case, any more 
> than they give a crap about the usecase of my laptop without vt tech. 
> (i.e. the removal of even the capability of using kqemu accelerator with 
> qemu)
>
>[...]
>
> Yes.  With the caveat - "when its time".  That time may be soas3, I 
> personally don't care.  I don't even have kids.  But there might be a 
> lot of people wanting soas on a 1G stick that would say- wait for soas4 
> or 5.
> 
> Admit at least that it is a real tradeoff.

You win, I admit it... No, wait... I don't:

The XO-1, with 1GB of flash, is perfectly capable of running Fedora 11
off an ext3 filesystem.

I bet we could easily drop plenty of bloat from SoaS without affecting
any user-visible feature, and fit nicely in a 1GB stick without fancy
filesystems. What do you think?


> liveusb-creator is the gui that provides the exact same functionality as 
> livecd-iso-to-stick, and runs on windows as well.  Were you not aware of 
> this?

I used it only once, but I had not understood that it was using
livecd-iso-to-stick under hood. Actually, it seems very hard to believe:
how can it possibly run such a Linux-dependent bash script under
Windows?


> > My compressed images came out a lot smaller than the ISO: 416MB vs
> > 589MB. Which, by itself, would make it the preferred distribution format
> > for many users with limited bandwidth.
> 
> when lzma is used for building the livecd iso, most or all of that win 
> is mute.

Indeed.

-- 
   // Bernie Innocenti - http://codewiz.org/
 \X/  Sugar Labs       - http://sugarlabs.org/