You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is the Fernet spec's versioning unambiguous? The spec on current mastersays:
This document describes version 0x80 (currently the only version) of the fernet format.
However, the current master is the sixteenth commit since the doc started, and it was called 0x80 when it started, too. What did those commits change? Should some of them have bumped the version?
The current lack of tags further suggests that version 0x80 is ambiguous.
From a skim of the 16 commits, here are the ones that jump out as potentially version-bumping changes:
ca8b345 pins to CBC, meaning that implementations of 1fa950a's 0x80 don't guarantee CBC.
58d2827 appears to change from nanoseconds to seconds—the commit message ("new format") and diff (wait—"gAAAAA" is new?! 😳 ) in 115850d further suggest a version-bump-worthy change.
3bbe466 "document checking version byte"—so it is new
c55d0d4 diff suggests that base64url was indicated in at least some specific locations in the original 0x80—was it so indicated everywhere?
In which case, the action item here is to ... what? Clean up Fernet's versioning somehow. By applying new version numbers to the version-bumping commits above? By dropping back to Git SHA for version? By adopting semantic versioning?
Activity
[-]is the version unambiguous?[/-][+]Is the version unambiguous?[/+]chadwhitacre commentedon May 4, 2016
From a skim of the 16 commits, here are the ones that jump out as potentially version-bumping changes:
ca8b345 pins to CBC, meaning that implementations of 1fa950a's 0x80 don't guarantee CBC.
58d2827 appears to change from nanoseconds to seconds—the commit message ("new format") and diff (wait—"gAAAAA" is new?! 😳 ) in 115850d further suggest a version-bump-worthy change.
3bbe466 "document checking version byte"—so it is new
c55d0d4 diff suggests that base64url was indicated in at least some specific locations in the original 0x80—was it so indicated everywhere?
0514d7d clear algorithm change
0250c59 Answer: no.
chadwhitacre commentedon May 4, 2016
The answer appears to be "yes."
In which case, the action item here is to ... what? Clean up Fernet's versioning somehow. By applying new version numbers to the version-bumping commits above? By dropping back to Git SHA for version? By adopting semantic versioning?
chadwhitacre commentedon May 5, 2016
To be explicit: this issues makes auditing somewhat more difficult.
chadwhitacre commentedon May 5, 2016
By tagging 0250c59 as 0x80 and adopting a versioning policy for future changes?