Skip to content

schema of timezones in the journal

We recently changed the format of new objects written in the journal: swh/devel/swh-docs!297 (closed)

However, offset_bytes is going to be renamed to simply offset in the model, as we are getting rid of the numeric offset. However, this will introduce yet another change to the journal format, and a somewhat confusing one.

What should we do about this?

  1. keep existing objects as-is, with a mix of offset_bytes, a numeric offset, and bytes offset
  2. keep the schema as it is currently, with a mix of offset_bytes and a numeric offset
  3. rewrite every object to use a bytes offset

Option 1 is the hardest on future consumers, and option 3 has the highest risk for corrupting existing data. Option 2 is a middle ground.


Migrated from T3890 (view on Phabricator)

Edited by Phabricator Migration user
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information