Hi Samir,Bron and I are discussing this as we speak. This feature will probably not be in the 3.x series, but in a new 4.x series. I hope to have some running code by the end of 2018. The ultimate goal is to make mailbox renames an atomic operation of nothing more than updating mailboxes.db entries in a single transaction. No moving of data on disk should be needed beyond updating cyrus.header with the current mailbox name. In other words, mailbox names will be reduced to meta-data.
On 07/20/2018 10:41 AM, Samir Aguiar wrote:
Hi all, We are planning to upgrade to cyrus 3.x.y and after going through the archives, I noticed there was some talk about using UUIDs as folder names for mailboxes[1][2]. However, after a few tests with 3.1.5 I noticed that the folder structure is still using mailbox names. #e4c9ac2bc[3] added the uniqueid parameter to `mboxname_datapath` (marked as unused) and after that there seems to be no further development on that side. Was this idea postponed? Is it still in the backlog or is it configurable? Sorry if I missed something, I'm a little late to the party. Thank you in advance. Kind regards, Samir Aguiar [1] https://lists.andrew.cmu.edu/pipermail/cyrus-devel/2015-March/003182.html [2] https://lists.andrew.cmu.edu/pipermail/cyrus-devel/2016-June/003786.html [3] https://github.com/cyrusimap/cyrus-imapd/commit/e4c9ac2bc88bc73fcc0f7151eb299ac1ed94333c#diff-3f7e9582b7051be83c0edd299be8f5a5
-- Ken Murchison Cyrus Development Team FastMail US LLC
<<attachment: murch.vcf>>
