#6269 NORM Never A: usb drive not recognized in journal in 690 when indexed by 670

[Zarro Boogs per Child]

#6269: usb drive not recognized in journal in 690 when indexed by 670
-------------------------------+--------------------------------------------
  Reporter:  erikos            |       Owner:  tomeu         
      Type:  defect            |      Status:  reopened      
  Priority:  normal            |   Milestone:  Never Assigned
 Component:  journal-activity  |     Version:                
Resolution:                    |    Keywords:                
  Verified:  0                 |    Blocking:                
 Blockedby:                    |  
-------------------------------+--------------------------------------------
Changes (by gnu):

  * status:  closed => reopened
  * resolution:  wontfix =>


Comment:

 I'm glad that the problem is temporarily reverted -- but the explanation
 points up a much more serious bug.  Next time we upgrade to a new xapian
 release, it will break.  This bug should stay open so that we fix xapian
 and/or Sugar *before* that happens.

 The proper fix is to never put crap "index" files onto other peoples' USB
 keys, nor should you try to read them from there.  Plugging in a
 filesystem on USB should not cause the OLPC, or Sugar, to start writing
 any files on the new filesystem!  And when writes occur, they should occur
 because the USER asked for something to be written there.  OLPC may own
 the NAND, but you don't own the USB media.  Didn't your momma teach you
 not to scribble in other peoples' books?

 But since OLPC has already determined not to fix the real problem, at
 least you could make the code that reads your crap files able to read the
 crap files created by *former versions of itself*!

 Of course, *forward* compatability is only half the problem, when dealing
 with portable media like USB sticks.  You would also need *backward*
 compatability, i.e. the new code should be writing a crap file that will
 work properly when the USB stick is inevitably plugged into older laptops.
 You don't need to imagine what'd happen in a school where some students
 are trying a new release while others continue on a more stable release:
 their USB sticks would stop working.  You have this bug report staring you
 in the face, but you closed it "wontfix".

 The crapfile reading and writing code should also be bulletproof in the
 face of *random binary gibberish* and *subtly altered text* fed to it in
 these crap files; read-only USB sticks, and completely full USB sticks,
 and almost-completely-full USB sticks that fill up while writing the crap
 files, and terabyte USB drives that would take hours to index but are
 removed before they finish; and other edge conditions.  If you insist on
 writing files in defiance of the user's wishes, the least you can do is
 make them reliable.

-- 
Ticket URL: <http://dev.laptop.org/ticket/6269#comment:2>
One Laptop Per Child <http://dev.laptop.org>
OLPC bug tracking system