On 13-08-09 9:33 PM, Andreas Gal wrote:
Ok. So neither require a flag day. Changing data formats might require
a flag day. But that would require a flag day in the couchdb world as
well. In other words the two flag day scenario is a red herring. Do
you agree?
No.
The picl-idp-backed auth story provides no mechanism for provisioning
new clients with full-strength keys of the kind Sync 1.1 expects. We
have no solution better than the existing PAKE pairing operation, which
is largely the impetus for what we are doing. Therefore new auth and
old auth clients are strictly separated. Since the user base most
likely to immediately want to use the new auth system are the people
already using Sync 1.1 (correct me if this assumption is wrong), and we
have no way to migrate them to the new auth system, there needs to be an
auth flag day.
In addition, Sync 1.1 expects records in a certain format, both envelope
and JSON payload. The existing Sync protocol and clients have no
support for negotiating upgrades to either format. We need to change at
least the JSON payload format for the next generation of Sync (and while
we're here, handle negotiating client format updates!). There will be a
flag day when old clients will not be able to sync against the new
record format.
We can have 2 flag days, or we can have 1, but I do not believe we can
have 0.
Nick
_______________________________________________
Sync-dev mailing list
[email protected]
https://mail.mozilla.org/listinfo/sync-dev
