[tahoe-lafs-trac-stream] [Tahoe-LAFS] #724: helper: client should check up on the helper's work

[Tahoe-LAFS]

#724: helper: client should check up on the helper's work
-------------------------------+---------------------------
     Reporter:  warner         |      Owner:
         Type:  enhancement    |     Status:  new
     Priority:  major          |  Milestone:  undecided
    Component:  code-encoding  |    Version:  1.4.1
   Resolution:                 |   Keywords:  upload-helper
Launchpad Bug:                 |
-------------------------------+---------------------------
Changes (by zooko):

 * keywords:  helper => upload-helper


Old description:

> Really, a client which uploads a file through the Helper should turn
> around immediately afterwards and verify that the helper really did it's
> job. The most intensive approach would be to perform a full verifier run
> (download all shares and check all bits). A more appropriate choice might
> be to perform a file check plus download the UEB from all shares to make
> sure it matches the value returned by the helper.
>
> Once #722 and #723 are fixed, a change like this would merely be a
> reliability improvement: if the helper is buggy or malicious (or not
> connected to the complete/correct serverlist), this would allow the
> client to detect it right away, while the original file is still
> available.
>
> The client would respond to a helper failure by performing a direct
> (helperless) upload, which is obviously slower but then removes reliance
> upon the helper for reliability/availability.

New description:

 Really, a client which uploads a file through the Helper should turn
 around immediately afterwards and verify that the helper really did it's
 job. The most intensive approach would be to perform a full verifier run
 (download all shares and check all bits). A more appropriate choice might
 be to perform a file check plus download the UEB from all shares to make
 sure it matches the value returned by the helper.

 Once #722 and #723 are fixed, a change like this would merely be a
 reliability improvement: if the helper is buggy or malicious (or not
 connected to the complete/correct serverlist), this would allow the client
 to detect it right away, while the original file is still available.

 The client would respond to a helper failure by performing a direct
 (helperless) upload, which is obviously slower but then removes reliance
 upon the helper for reliability/availability.

--

--
Ticket URL: <https://tahoe-lafs.org/trac/tahoe-lafs/ticket/724#comment:1>
Tahoe-LAFS <https://Tahoe-LAFS.org>
secure decentralized storage