[Sugar-devel] [IAEP] SLOBs Position on SoaS

[Douglas McClendon]

Bill Bogstad wrote:
> On Wed, Sep 16, 2009 at 8:22 PM, Douglas McClendon
> <dmc.sugar at filteredperception.org> wrote:
>> Bill Bogstad wrote:
>>> ...
>>>
>>> I also don't think we can leave Sugar LiveUSB to any distribution.
>>> My impression is that both LiveCD and LiveUSB Linux distributions are
>>> essentially gimmicks for all of them.
>> I generally agree with the rest of your sentiments in this mail, but as the
>> now quasi-official 'Godfather of Fedora LiveCD' I have to respond to this
>> 'gimmick' claim.
> 
> I'm sorry if you were offended by the word gimmick.  A better word
> might have been niche.  From the first time, I saw Knoppix  Live media
> system were clearly interesting and useful.  However...

I'm not offended, I was just feeling old and like giving a lecture on 
the dark ol' days of linux a decade ago.

> 
>> In summary - LiveUSB == primarily trial and installation medium.  I.e.
>> perhaps the thing that _generates_
>> 'installed-normal-nonlive-fedora-on-a-stick' on sticks whose flash is
>> performancewise on par or better than a usb rotating disk.
> 
> Even you don't see this as a normal everyday usage model. 

No, I can totally see even soas-v1 as it is being useful everyday for a 
fairly significant audience.  If that audience can be grown much further 
with various enhancements or alternate approaches, thats all the better.

  Assuming
> that to be widely used, Sugar needs to support this model where do we
> go from here?  Should Sugar wait for distributions to make it more
> usable?  Am I wrong about this being vital to Sugar success?  Linux
> itself got its start coming in the backdoor.  Given the lack of
> marketing dollars available to Sugar, I think a similar strategy is
> worth considering.
> 
> You second message about changing flash/filesystem technologies brings
> to mind a discussion on a different mailing list about whether SSDs
> are appropriate for use as journals for enterprise databases.  Many
> people are finding that they see great performance improvements.
> There are quite a few netbooks out there at this point and I haven't
> heard anything about massive flash failures at those price points
> either.   I'm wondering if people are just scared of the fact that
> flash is different then hard drives.   The differences aren't all bad
> either.  Flash doesn't suffer from head crashes where you lose read
> access to ALL your data.  Instead, you probably slowly lose the
> ability to change it.  Somehow that sounds better to me.

I can easily point to two anecdotal experiences of exactly the opposite 
cases, but I'll generally cede to your more relevant experience.  I.e. 
in addition to countless* seemingly unfixably corrupted filesystems on 
usbsticks, I just had a 2G microSD 100% instantaneously just stop 
responding (no reader recognizes it).  And 10+ years ago as a sysadmin 
for 20+ engineering R&D group, I had an awesome core server 9G disk 
degrade slowly to 1/3MB/s read speed, and had fun watching it slowly 
replicate, instead of having to go to the nightly backup instead.

Those were fun pre-9/11 times...

-dmc

* I am a developer, so when I repeatedly run into the same failure, and 
don't try too hard to avoid it, it doesn't represent the same usage 
experience of an end-user.


> 
> When Sugar was being run an XO which had its flash soldered to the
> motherboard, it made sense to care alot about reducing flash writes.
> If SoaS is deployed with integrated XS backups of user files, so what
> if the USB flash drive only lasts a year or two.  It's still way
> cheaper then the cost of an XO. [This presupposes some actual testing
> to determine what the typical lifespan would be.]
> 
> Bill Bogstad