[tahoe-lafs-trac-stream] [Tahoe-LAFS] #3915: Size passed to IStorageServer.allocate_bytes is bigger than actual share size

[Tahoe-LAFS]

#3915: Size passed to IStorageServer.allocate_bytes is bigger than actual share
size
----------------------+---------------------------------------
 Reporter:  itamarst  |          Owner:
     Type:  defect    |         Status:  new
 Priority:  normal    |      Milestone:  HTTP Storage Protocol
Component:  unknown   |        Version:  n/a
 Keywords:            |  Launchpad Bug:
----------------------+---------------------------------------
 `immutable/layout.py` calculates share size using the _maximum_ possible
 URI extension size. The actual share has a specific, typically smaller URI
 extension size (as generated in `immutable/encode.py`). So
 `IBucketWriter.close()` is called after sending fewer bytes than were
 claimed in `allocate_buckets()`.

 For Foolscap this works, however ugly it is. For current HTTP storage
 protocol, this does not.

 Instead of fixing the HTTP protocol to have a notion of "closing early",
 seems better to just fix the size calculation so it uses the actual size
 of the URI extension.

 Note that there may be additional sources of discrepancy :(

--
Ticket URL: <https://tahoe-lafs.org/trac/tahoe-lafs/ticket/3915>
Tahoe-LAFS <https://Tahoe-LAFS.org>
secure decentralized storage